iii

The Gauge Block Handbook

by

Ted Doiron and John Beers

Dimensional Metrology GroupPrecision Engineering Division

National Institute of Standards and Technology

Preface

The Dimensional Metrology Group, and its predecessors at the National Institute of Standardsand Technology (formerly the National Bureau of Standards) have been involved indocumenting the science of gauge block calibration almost continuously since the seminal workof Peters and Boyd in 1926 [1]. Unfortunately, most of this documentation has been in the formof reports and other internal documents that are difficult for the interested metrologist outside theInstitute to obtain.

On the occasion of the latest major revision of our calibration procedures we decided to assembleand extend the existing documentation of the NIST gauge block calibration program into onedocument. We use the word assemble rather than write because most of the techniques describedhave been documented by various members of the Dimensional Metrology Group over the last 20years. Unfortunately, much of the work is spread over multiple documents, many of the detailsof the measurement process have changed since the publications were written, and many largegaps in coverage exist. It is our hope that this handbook has assembled the best of the previousdocumentation and extended the coverage to completely describe the current gauge blockcalibration process.

Many of the sections are based on previous documents since very little could be added incoverage. In particular, the entire discussion of single wavelength interferometry is due to JohnBeers [2]; the section on preparation of gauge blocks is due to Clyde Tucker [3]; the section onthe mechanical comparator techniques is predominantly from Beers and Tucker [4]; and theappendix on drift eliminating designs is an adaptation for dimensional calibrations of the work ofJoseph Cameron [5] on weighing designs. They have, however, been rewritten to make thehandbook consistent in style and coverage. The measurement assurance program has beenextensively modified over the last 10 years by one of the authors (TD), and chapter 4 reflectsthese changes.

iv

We would like to thank Mr. Ralph Veale, Mr. John Stoup, Mrs. Trish Snoots, Mr. Eric Stanfield,Mr. Dennis Everett, Mr. Jay Zimmerman, Ms. Kelly Warfield and Dr. Jack Stone, the membersof the Dimensional Metrology Group who have assisted in both the development and testing ofthe current gauge block calibration system and the production of this document.

TD and JSB

v

CONTENTS

PagePreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1. Length

1.1 The meter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 The inch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2. Gauge blocks

2.1 A short history of gauge blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.2 Gauge block standards (U.S.). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.2.1 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.2.2 Nomenclature and definitions . . . . . . . . . . . . . . . . . . . . . . . 5

2.2.3 Tolerance grades . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.2.4 Recalibration requirements . . . . . . . . . . . . . . . . . . . . . . . . . 10

2.3 International standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3. Physical and thermal properties of gauge blocks

3.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.2 Flatness and parallelism

3.2.1 Flatness measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.2.2 Parallelism measurement . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.3 Thermal expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3.3.1 Thermal expansion of gauge block materials . . . . . . . . . . . 19

3.3.2 Thermal expansion uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.3 Elastic properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.4.1 Contact deformation in mechanical comparisons . . . . . . . 26

3.4.2 Measurement of probe force and tip radius . . . . . . . . . . . . 28

3.5 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

vi

4. Measurement assurance programs

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4.2 A comparison: traditional metrology vs measurement assurance programs

4.2.1 Tradition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... . . . . . . . 33

4.2.2 Process control: a paradigm shift . . . . . . . . . . . . . . . . . . . . 34

4.2.3 Measurement assurance: building a measurement process model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

4.3 Determining Uncertainty

4.3.1 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4.3.2 Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

4.3.3 Random error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.3.4 Systematic error and type B uncertainty . . . . . . . . . . . . . . 41

4.3.5 Error budgets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

4.3.6 Combining type A and type B uncertainties . . . . . . . . . . . . 46

4.3.7 Combining random and systematic errors . . . . . . . . . . . . . . . . . . . . 47

4.4 The NIST gauge block measurement assurance program

4.4.1 Establishing interferometric master values . . . . . . . . . . . . . 49

4.4.2 The comparison process . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4.4.2.1 Measurement schemes - drift eliminating designs . . . . . . . . . . . . . . . . . . . . 53

4.4.2.2 Control parameter for repeatability . . . . . . . . 56

4.4.2.3 Control test for variance . . . . . . . . . . . . . . . . . 59

4.4.2.4 Control parameter (S-C). . . . . . . . . . . . . . . . . 60

4.4.2.5 Control test for (S-C), the check standard . . . 63

4.4.2.6 Control test for drift . . . . . . . . . . . . . . . . . . . . . 64

4.4.3 Calculating total uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . 64

4.5 Summary of the NIST measurement assurance program . . . . . . . . . . . 66

vii

5. The NIST mechanical comparison procedure

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

5.2 Preparation and inspection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

5.2.1 Cleaning procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

5.2.2 Cleaning interval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

5.2.3 Storage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

5.2.4 Deburring gauge blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

5.3 The comparative principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

5.3.1 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

5.4 Gauge block comparators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

5.4.1 Scale and contact force control . . . . . . . . . . . . . . . . . . . . . . 75

5.4.2 Stylus force and penetration corrections . . . . . . . . . . . . . . . 76

5.4.3 Environmental factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

5.4.3.1 Temperature effects . . . . . . . . . . . . . . . . . . . . 78

5.4.3.2 Control of temperature effects . . . . . . . . . . . . . . . . . 80

5.5 Intercomparison procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

5.5.1 Handling techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

5.6 Comparison designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

5.6.1 Drift eliminating designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

5.6.1.1 The 12/4 design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

5.6.1.2 The 6/3 design . . . . . . . . . . . . . . . . . . . . . . . . 84

5.6.1.3 The 8/4 design . . . . . . . . . . . . . . . . . . . . . . . . 85

5.6.1.4 The ABBA design . . . . . . . . . . . . . . . . . . . . . . 86

5.6.2 Example of calibration output using the 12/4 design . . . . . . 86

5.7 Current NIST system performance . . . . . . . . . . . . . . . . . . . . . . . . . 88

5.7.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

6. Gauge block interferometry

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

6.2 Interferometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

6.2.1 The Kosters type interferometer . . . . . . . . . . . . . . . . . . . . . . 93

6.2.2 The NPL interferometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

6.2.3 Testing optical quality of interferometers . . . . . . . . . . . . . . . 97

6.2.4 Interferometer corrections . . . . . . . . . . . . . . . . . . . . . . . . . . 98

6.2.5 Laser light sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

6.3 Environmental conditions and their measurement . . . . . . . . . . . . . 102

6.3.1 Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

6.3.2 Atmospheric pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

6.3.3 Water vapor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

6.4 Gauge block measurement procedure . . . . . . . . . . . . . . . . . . . . . . 105

6.5 Computation of gauge block length . . . . . . . . . . . . . . . . . . . . . . . . . 107

6.5.1 Calculation of the wavelength . . . . . . . . . . . . . . . . . . . . . . . 107

6.5.2 Calculation of the whole number of fringes . . . . . . . . . . . . . 109

6.5.3 Calculation of the block length from observed data . . . . . . . 109

6.6 Type A and B errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

6.7 Process evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

6.8 Multiple wavelength interferometry . . . . . . . . . . . . . . . . . . . . . . . . . 115

6.9 Use of the line scale interferometer for end standard calibration . . 117

7. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

Appendix A. Drift eliminating designs for non-simultaneous comparison calibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

Appendix B. Wringing films . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138Appendix C. Phase shifts in gauge block interferometry . . . . . . . . . . . . . . . . . . . 141Appendix D. Deformation corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

viii

1

Gauge Block Handbook

Introduction

Gauge block calibration is one of the oldest high precision calibrations made in dimensionalmetrology. Since their invention at the turn of the century gauge blocks have been the majorsource of length standardization for industry. In most measurements of such enduringimportance it is to be expected that the measurement would become much more accurate andsophisticated over 80 years of development. Because of the extreme simplicity of gauge blocksthis has only been partly true. The most accurate measurements of gauge blocks have notchanged appreciably in accuracy in the last 70 years. What has changed is the much morewidespread necessity of such accuracy. Measurements, which previously could only be madewith the equipment and expertise of a national metrology laboratory, are routinely expected inprivate industrial laboratories.

To meet this widespread need for higher accuracy, the calibration methods used for gauge blockshave been continuously upgraded. This handbook is a both a description of the current practiceat the National Institute of Standards and Technology, and a compilation of the theory and lore ofgauge block calibration. Most of the chapters are nearly self-contained so that the interestedreader can, for example, get information on the cleaning and handling of gauge blocks withouthaving to read the chapters on measurement schemes or process control, etc. This partitioning ofthe material has led to some unavoidable repetition of material between chapters.

The basic structure of the handbook is from the theoretical to the practical. Chapter 1 concernsthe basic concepts and definitions of length and units. Chapter 2 contains a short history ofgauge blocks, appropriate definitions and a discussion of pertinent national and internationalstandards. Chapter 3 discusses the physical characteristics of gauge blocks, including thermal,mechanical and optical properties. Chapter 4 is a description of statistical process control (SPC)and measurement assurance (MA) concepts. The general concepts are followed by details of theSPC and MA used at NIST on gauge blocks.

Chapters 5 and 6 cover the details of the mechanical comparisons and interferometric techniquesused for gauge block calibrations. Full discussions of the related uncertainties and correctionsare included. Finally, the appendices cover in more detail some important topics in metrologyand gauge block calibration.

2

1.0 Length

1.1 The Meter

At the turn of 19th century there were two distinct major length systems. The metric length unitwas the meter that was originally defined as 1/10,000,000 of the great arc from the pole to theequator, through Paris. Data from a very precise measurement of part of that great arc was usedto define an artifact meter bar, which became the practical and later legal definition of the meter. The English system of units was based on a yard bar, another artifact standard [6].

These artifact standards were used for over 150 years. The problem with an artifact standard forlength is that nearly all materials are slightly unstable and change length with time. For example,by repeated measurements it was found that the British yard standard was slightly unstable. Theconsequence of this instability was that the British inch ( 1/36 yard) shrank [7], as shown in table1.1.

Table 1.1

1895 - 25.399978 mm

1922 - 25.399956 mm

1932 - 25.399950 mm

1947 - 25.399931 mm

The first step toward replacing the artifact meter was taken by Albert Michelson, at the request ofthe International Committee of Weights and Measures (CIPM). In 1892 Michelson measured themeter in terms of the wavelength of red light emitted by cadmium. This wavelength was chosenbecause it has high coherence, that is, it will form fringes over a reasonable distance. Despite thework of Michelson, the artifact standard was kept until 1960 when the meter was finallyredefined in terms of the wavelength of light, specifically the red-orange light emitted by excitedkrypton-86 gas.

Even as this definition was accepted, the newly invented helium-neon laser was beginning to beused for interferometry. By the 1970's a number of wavelengths of stabilized lasers wereconsidered much better sources of light than krypton red-orange for the definition of the meter. Since there were a number of equally qualified candidates the International Committee onWeights and Measures (CIPM) decided not to use any particular wavelength, but to make achange in the measurement hierarchy. The solution was to define the speed of light in vacuum asexactly 299,792,458 m/s, and make length a derived unit. In theory, a meter can be produced byanyone with an accurate clock [8].

3

In practice, the time-of-flight method is impractical for most measurements, and the meter ismeasured using known wavelengths of light. The CIPM lists a number of laser and atomicsources and recommended frequencies for the light. Given the defined speed of light, thewavelength of the light can be calculated, and a meter can be generated by counting wavelengthsof the light. Methods for this measurement are discussed in the chapter on interferometry.

1.2 The Inch

In 1866, the United Stated Surveyor General decided to base all geodetic measurements on aninch defined from the international meter. This inch was defined such that there were exactly39.37 inches in the meter. England continued to use the yard bar to define the inch. Thesedifferent inches continued to coexist for nearly 100 years until quality control problems duringWorld War II showed that the various inches in use were too different for completelyinterchangeable parts from the English speaking nations. Meetings were held in the 1950's andin 1959 the directors of the national metrology laboratories of the United States, Canada,England, Australia and South Africa agreed to define the inch as 25.4 millimeters, exactly [9]. This definition was a compromise; the English inch being somewhat longer, and the U.S. inchsmaller. The old U.S. inch is still in use for commercial surveying of land in the form of the"surveyor's foot," which is 12 old U.S. inches.

4

2.0 Gauge Blocks

2.1 A Short History of Gauge Blocks

By the end of the nineteenth century the idea of interchangeable parts begun by Eli Whitney hadbeen accepted by industrial nations as the model for industrial manufacturing. One of thedrawbacks to this new system was that in order to control the size of parts numerous gauges wereneeded to check the parts and set the calibrations of measuring instruments. The number ofgauges needed for complex products, and the effort needed to make and maintain the gauges wasa significant expense. The major step toward simplifying this situation was made by C.E.Johannson, a Swedish machinist.

Johannson's idea, first formulated in 1896 [10], was that a small set of gauges that could becombined to form composite gauges could reduce the number of gauges needed in the shop. Forexample, if four gauges of sizes 1 mm, 2 mm, 4 mm, and 8 mm could be combined in anycombination, all of the millimeter sizes from 1 mm to 15 mm could be made from only these fourgauges. Johannson found that if two opposite faces of a piece of steel were lapped very flat andparallel, two blocks would stick together when they were slid together with a very small amountof grease between them. The width of this "wringing" layer is about 25 nm, and was so small forthe tolerances needed at the time, that the block lengths could be added together with nocorrection for interface thickness. Eventually the wringing layer was defined as part of thelength of the block, allowing the use of an unlimited number of wrings without correction for thesize of the wringing layer.

In the United States, the idea was enthusiastically adopted by Henry Ford, and from his examplethe use of gauge blocks was eventually adopted as the primary transfer standard for length inindustry. By the beginning of World War I, the gauge block was already so important to industrythat the Federal Government had to take steps to insure the availability of blocks. At theoutbreak of the war, the only supply of gauge blocks was from Europe, and this supply wasinterrupted.

In 1917 inventor William Hoke came to NBS proposing a method to manufacture gauge blocksequivalent to those of Johannson [11]. Funds were obtained from the Ordnance Department forthe project and 50 sets of 81 blocks each were made at NBS. These blocks were cylindrical andhad a hole in the center, the hole being the most prominent feature of the design. The currentgeneration of square cross-section blocks have this hole and are referred to as "Hoke blocks."

2.2 Gauge Block Standards (U.S.)

There are two main American standards for gauge blocks, the Federal Specification GGG-G-15C[12] and the American National Standard ANSI/ASME B89.1.9M [13]. There are very fewdifferences between these standards, the major ones being the organization of the material andthe listing of standard sets of blocks given in the GGG-G-15C specification. The material in theASME specification that is pertinent to a discussion of calibration is summarized below.

5

2.2.1 Scope

The ASME standard defines all of the relevant physical properties of gauge blocks up to 20inches and 500 mm long. The properties include the block geometry (length, parallelism,flatness and surface finish), standard nominal lengths, and a tolerance grade system forclassifying the accuracy level of blocks and sets of blocks.

The tolerancing system was invented as a way to simplify the use of blocks. For example,suppose gauge blocks are used to calibrate a certain size fixed gauge, and the required accuracyof the gauge is 0.5 µm. If the size of the gauge requires a stack of five blocks to make up thenominal size of the gauge the accuracy of each block must be known to 0.5/5 or 0.1 µm. This isnear the average accuracy of an industrial gauge block calibration, and the tolerance could bemade with any length gauge blocks if the calibrated lengths were used to calculate the length ofthe stack. But having the calibration report for the gauge blocks on hand and calculating thelength of the block stack are a nuisance. Suppose we have a set of blocks which are guaranteedto have the property that each block is within 0.05 µm of its nominal length. With thisknowledge we can use the blocks, assume the nominal lengths and still be accurate enough forthe measurement.

The tolerance grades are defined in detail in section 2.2.3, but it is important to recognize thedifference between gauge block calibration and certification. At NIST, gauge blocks arecalibrated, that is, the measured length of each block is reported in the calibration report. Thereport does not state which tolerance grade the blocks satisfy. In many industrial calibrationsonly the certified tolerance grade is reported since the corrections will not be used.

2.2.2 Nomenclature and Definitions

A gauge block is a length standard having flat and parallel opposing surfaces. The cross-sectional shape is not very important, although the standard does give suggested dimensions forrectangular, square and circular cross-sections. Gauge blocks have nominal lengths defined ineither the metric system (millimeters) or in the English system (1 inch = 25.4 mm).

The length of the gauge block is defined at standard reference conditions:

temperature = 20 ºC (68 ºF )barometric pressure = 101,325 Pa (1 atmosphere)water vapor pressure = 1,333 Pa (10 mm of mercury)CO2 content of air = 0.03%.

Of these conditions only the temperature has a measurable effect on the physical length of theblock. The other conditions are needed because the primary measurement of gauge block lengthis a comparison with the wavelength of light. For standard light sources the frequency of thelight is constant, but the wavelength is dependent on the temperature, pressure, humidity, andCO2 content of the air. These effects are described in detail later.

6

The length of a gauge block is defined as the perpendicular distance from a gauging point on oneend of the block to an auxiliary true plane wrung to the other end of the block, as shown in figure2.1 (from B89.1.9).

Figure 2.1. The length of a gauge block is the distance from the gauging point on the top surface to the plane of the platen adjacent to the wrung gaugeblock.

This length is measured interferometrically, as described later, and corrected to standard conditions.It is worth noting that gauge blocks are NEVER measured at standard conditions because thestandard vapor pressure of water of 10 mm of mercury is nearly 60% relative humidity that wouldallow steel to rust. The standard conditions are actually spectroscopic standard conditions, i.e., theconditions at which spectroscopists define the wavelengths of light.

This definition of gauge block length that uses a wringing plane seems odd at first, but is veryimportant for two reasons. First, light appears to penetrate slightly into the gauge block surface, aresult of the surface finish of the block and the electromagnetic properties of metals. If the wringingplane and the gauge block are made of the same material and have the same surface finish, then thelight will penetrate equally into the block top surface and the reference plane, and the errors cancel. If the block length was defined as the distance between the gauge block surfaces the penetrationerrors would add, not cancel, and the penetration would have to be measured so a correction couldbe made. These extra measurements would, of course, reduce the accuracy of the calibration.

The second reason is that in actual use gauge blocks are wrung together. Suppose the length ofgauge blocks was defined as the actual distance between the two ends of the gauge block, not wrungto a plane. For example, if a length of 6.523 mm is needed gauge blocks of length 2.003 mm, 2.4

widthdepth

auxiliary platel gl

7

mm, and 2.12 mm are wrung together. The length of this stack is 6.523 plus the length of twowringing layers. It could also be made using the set (1 mm, 1 mm, 1 mm, 1.003 mm, 1.4 mm, and1.12 mm) which would have the length of 6.523 mm plus the length of 5 wringing layers. In orderto use the blocks these wringing layer lengths must be known. If, however, the length of each blockcontains one wringing layer length then both stacks will be of the same defined length.

NIST master gauge blocks are calibrated by interferometry in accordance with the definition of gaugeblock length. Each master block carries a wringing layer with it, and this wringing layer istransferred to every block calibrated at NIST by mechanical comparison techniques.

The mechanical length of a gauge block is the length determined by mechanical comparison of ablock to another block of known interferometrically determined length. The mechanical comparisonmust be a measurement using two designated points, one on each end of the block. Since most gaugeblock comparators use mechanical contact for the comparison, if the blocks are not of the samematerial corrections must be made for the deformation of the blocks due to the force of thecomparator contact.

The reference points for rectangular blocks are the center points of each gauging face. For squaregauge block mechanical comparison are shown in figure 2.2.

Figure 2.2 Definition of the gauging point on square gauge blocks.

For rectangular and round blocks the reference point is the center of gauging face. For round orsquare blocks that have a center hole, the point is midway between the hole edge and the edge of theblock nearest to the size marking.

1/2 distance between edge ofblock and edge of countersink

1/2 width

25 m

m

8

2.2.3 Tolerance Grades

There are 4 tolerance grades; 0.5, 1, 2, and 3. Grades 0.5 and 1 gauge blocks have lengths very closeto their nominal values. These blocks are generally used as calibration masters. Grades 2 and 3 areof lower quality and are used for measurement and gauging purposes. Table 2.1 shows the length,flatness and parallelism requirements for each grade. The table shows that grade 0.5 blocks arewithin 1 millionth of an inch (1 µin) of their nominal length, with grades 1, 2, and 3 each roughlydoubling the size of the maximum allowed deviation.

Table 2.1a Tolerance Grades for Inch Blocks (in µin )

Nominal Grade .5 Grade 1 Grade 2 Grade 3

<1 inch 1 2 +4, -2 +8, -4 2 2 4 +8, -4 +16, -8 3 3 5 +10, -5 +20, -10 4 4 6 +12, -6 +24, -12 5 7 +14, -7 +28, -14 6 8 +16, -8 +32, -16 7 9 +18, -9 +36, -18 8 10 +20, -10 +40, -2010 12 +24, -12 +48, -2412 14 +28, -14 +56, -2816 18 +36, -18 +72, -3620 20 +40, -20 +80, -40

Table 2.1b Tolerance Grades for Metric Blocks ( µm )

Nominal Grade .5 Grade 1 Grade 2 Grade 3

< 10 mm 0.03 0.05 +0.10, -0.05 +0.20, -0.10< 25 mm 0.03 0.05 +0.10, -0.05 +0.30, -0.15< 50 mm 0.05 0.10 +0.20, -0.10 +0.40, -0.20< 75 mm 0.08 0.13 +0.25, -0.13 +0.45, -0.23< 100 mm 0.10 0.15 +0.30, -0.15 +0.60, -0.30 125 mm 0.18 +0.36, -0.18 +0.70, -0.35 150 mm 0.20 +0.41, -0.20 +0.80, -0.40

175 mm 0.23 +0.46, -0.23 +0.90, -0.45 200 mm 0.25 +0.51, -0.25 +1.00, -0.50

250 mm 0.30 +0.60, -0.30 +1.20, -0.60 300 mm 0.35 +0.70, -0.35 +1.40, -0.70 400 mm 0.45 +0.90, -0.45 +1.80, -0.90 500 mm 0.50 +1.00, -0.50 +2.00, -1.90

9

Since there is uncertainty in any measurement, the standard allows for an additional tolerance forlength, flatness, and parallelism. These additional tolerances are given in table 2.2.

Table 2.2 Additional Deviations for Measurement Uncertainty

Nominal Grade .5 Grade 1 Grade 2 Grade 3 in (mm) µin (µm) µin (µm) µin (µm) µin (µm)

< 4 (100) 1 (0.03) 2 (0.05) 3 (0.08) 4 (0.10)< 8 (200) 3 (0.08) 6 (0.15) 8 (0.20)< 12 (300) 4 (0.10) 8 (0.20) 10 (0.25)< 20 (500) 5 (0.13) 10 (0.25) 12 (0.30)

For example, for a grade 1 gauge block of nominally 1 inch length the length tolerance is 2 µin. With the additional tolerance for measurement uncertainty from table 2 of 2 uin, a 1 in grade 1 blockmust have a measured length within 4 uin of nominal.

2.2.4 Recalibration Requirements

There is no required schedule for recalibration of gauge blocks, but both the ASME and Federalstandards recommend recalibration periods for each tolerance grade, as shown below:

Grade Recalibration Period 0.5 Annually 1 Annually 2 Monthly to semi-annually

3 Monthly to semi-annually

Most NIST customers send master blocks for recalibration every 2 years. Since most master blocksare not used extensively, and are used in a clean, dry environment this schedule is probably adequate. However, despite popular misconceptions, NIST has no regulatory power in these matters. Therules for tolerance grades, recalibration, and replacement rest entirely with the appropriategovernment agency inspectors.

2.3 International Standards

Gauge blocks are defined internationally by ISO Standard 3650 [14], the current edition being thefirst edition 1978-07-15. This standard is much like the ANSI standard in spirit, but differs in mostdetails, and of course does not define English size blocks.

10

The length of the gauge block is defined as the distance between a flat surface wrung to one end ofthe block, and a gauging point on the opposite end. The ISO specification only defines rectangularcross-sectioned blocks and the gauging point is the center of the gauging face. The non-gaugingdimensions of the blocks are somewhat smaller than the corresponding ANSI dimensions.

There are four defined tolerance grades in ISO 3650; 00, 0, 1 and 2. The algorithm for the lengthtolerances are shown in table 2.3, and there are rules for rounding stated to derive the tables includedin the standard.

Table 2.3

Grade Deviation from NominalLength (µm)

00 (0.05 + 0.0001L) 0 (0.10 + 0.0002L) 1 (0.20 + 0.0004L) 2 (0.40 + 0.0008L)

Where L is the block nominal length in millimeters.

The ISO standard does not have an added tolerance for measurement uncertainty; however, the ISOtolerances are comparable to those of the ANSI specification when the additional ANSI tolerancefor measurement uncertainty is added to the tolerances of Table 2.1.

11

Figure 2.3. Comparison of ISO grade tolerances(black dashed) and ASME grade tolerances (red).

A graph of the length tolerance versus nominal length is shown in figure 2.3. The different classtolerance for ISO and ANSI do not match up directly. The ANSI grade 1 is slightly tighter than ISOclass 00, but if the additional ANSI tolerance for measurement uncertainty is used the ISO Grade 00is slightly tighter. The practical differences between these specifications are negligible.

In many countries the method for testing the variation in length is also standardized. For example,in Germany [15] the test block is measured in 5 places: the center and near each corner (2 mm fromeach edge). The center gives the length of the block and the four corner measurements are used tocalculate the shortest and longest lengths of the block. Some of the newer gauge block comparatorshave a very small lower contact point to facilitate these measurements very near the edge of theblock.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

0 100 200 300 400 500

millimeters

mic

rom

eter

sISO 2

ISO K,1

ISO 0US 3

ISO 00US 2

US 1

12

3. Physical and Thermal Properties of Gauge Blocks

3.1 Materials

From the very beginning gauge blocks were made of steel. The lapping process used to finish theends, and the common uses of blocks demand a hard surface. A second virtue of steel is that mostindustrial products are made of steel. If the steel gauge block has the same thermal expansioncoefficient as the part to be gauged, a thermometer is not needed to obtain accurate measurements. This last point will be discussed in detail later.

The major problem with gauge blocks was always the stability of the material. Because of thehardening process and the crystal structure of the steel used, most blocks changed length in time. For long blocks, over a few inches, the stability was a major limitation. During the 1950s and 1960sa program to study the stability problem was sponsored by the National Bureau of Standards and theASTM [16,17]. A large number of types of steel and hardening processes were tested to discovermanufacturing methods that would produce stable blocks. The current general use of 52100hardened steel is the product of this research. Length changes of less than 1 part in 10-6/decade arenow common.

Over the years, a number of other materials were tried as gauge blocks. Of these, tungsten carbide,chrome carbide, and Cervit are the most interesting cases.

The carbide blocks are very hard and therefore do not scratch easily. The finish of the gaugingsurfaces is as good as steel, and the lengths appear to be at least as stable as steel, perhaps even morestable. Tungsten carbide has a very low expansion coefficient (1/3 of steel) and because of the highdensity the blocks are deceptively heavy. Chrome carbide has an intermediate thermal expansioncoefficient (2/3 of steel) and is roughly the same density as steel. Carbide blocks have become verypopular as master blocks because of their durability and because in a controlled laboratoryenvironment the thermal expansion difference between carbide and steel is easily manageable.

Cervit is a glassy ceramic that was designed to have nearly zero thermal expansion coefficient. Thisproperty, plus a zero phase shift on quartz platens (phase shift will be discussed later), made thematerial attractive for use as master blocks. The drawbacks are that the material is softer than steel,making scratches a danger, and by nature the ceramic is brittle. While a steel block might bedamaged by dropping, and may even need stoning or recalibration, Cervit blocks tended to crack orchip. Because the zero coefficient was not always useful and because of the combination of softnessand brittleness they never became popular and are no longer manufactured.

A number of companies are experimenting with zirconia based ceramics, and one type is beingmarketed. These blocks are very hard and have thermal expansion coefficient of approximately 9x 10-6/ºC, about 20% lower than steel.

13

3.2 Flatness and Parallelism

We will describe a few methods that are useful to characterize the geometry of gauge blocks. It isimportant to remember, however, that these methods provide only a limited amount of data aboutwhat can be, in some cases, a complex geometric shape. When more precise measurements or apermanent record is needed, the interference fringe patterns can be photographed. The usefulnessof each of the methods must be judged in the light of the user's measurement problem.

3.2.1 Flatness Measurements.

Various forms of interferometers are applicable to measuring gauge block flatness. All produceinterference fringe patterns formed with monochromatic light by the gauge block face and areference optical flat of known flatness. Since modest accuracies (25 nm or 1 µin) are generallyneeded, the demands on the light source are also modest. Generally a fluorescent light with a greenfilter will suffice as an illumination source. For more demanding accuracies, a laser or atomicspectral lamp must be used.

The reference surface must satisfy two requirements. First, it must be large enough to cover theentire surface of the gauge block. Usually a 70 mm diameter or larger is sufficient. Secondly, thereference surface of the flat should be sufficiently planar that any fringe curvature can be attributedsolely to the gauge block. Typical commercially available reference flats, flat to 25 nm over a70 mm diameter, are usually adequate.

Gauge blocks 2 mm (0.1 in.) and greater can be measured in a free state, that is, not wrung to aplaten. Gauge blocks less than 2 mm are generally flexible and have warped surfaces. There is nocompletely meaningful way to define flatness. One method commonly used to evaluate the"flatness" is by "wringing" the block to another more mechanically stable surface. When the blockis wrung to the surface the wrung side will assume the shape of the surface, thus this surface will beas planar as the reference flat.

We wring these thin blocks to a fused silica optical flat so that the wrung surface can be viewedthrough the back surface of the flat. The interface between the block and flat, if wrung properly,should be a uniform gray color. Any light or colored areas indicate poor wringing contact that willcause erroneous flatness measurements. After satisfactory wringing is achieved the upper (non-wrung) surface is measured for flatness. This process is repeated for the remaining surface of theblock.

Figures 3.1a and 3.1b illustrate typical fringe patterns. The angle between the reference flat andgauge block is adjusted so that 4 or 5 fringes lie across the width of the face of the block, as in figure3.1a, or 2 or 3 fringes lie along the length of the face as in figure 3.1b. Four fringes in each directionare adequate for square blocks.

14

Figure 3.1 a, b, and c. Typical fringe patterns used to measure gauge block flatness. Curvature canbe measured as shown in the figures.

The fringe patterns can be interpreted as contour maps. Points along a fringe are points of equalelevation and the amount of fringe curvature is thus a measure of planarity.

Curvature = a/b (in fringes).

For example, a/b is about 0.2 fringe in figure 3.1a and 0.6 fringe in figure 3.1b. Conversion tolength units is accomplished using the known wavelength of the light. Each fringe represents a one-half wavelength difference in the distance between the reference flat and the gauge block. Greenlight is often used for flatness measurements. Light in the green range is approximately 250 nm(10 µin ) per fringe, therefore the two illustrations indicate flatness deviations of 50 nm and 150 nm(2 µin and 6 µin ) respectively.

Another common fringe configuration is shown in figure 3.1c. This indicates a twisted gauging face. It can be evaluated by orienting the uppermost fringe parallel to the upper gauge block edge and thenmeasuring "a" and "b" in the two bottom fringes. the magnitude of the twist is a/b which in this caseis 75 nm (3 µin) in green.

In manufacturing gauge blocks, the gauging face edges are slightly beveled or rounded to eliminatedamaging burrs and sharpness. Allowance should be made for this in flatness measurements byexcluding the fringe tips where they drop off at the edge. Allowances vary, but 0.5 mm ( 0.02 in)is a reasonable bevel width to allow.

3.2.2 Parallelism measurement

Parallelism between the faces of a gauge block can be measured in two ways; with interferometryor with an electro-mechanical gauge block comparator.

a

b

ab

a b

15

Interferometer Technique

The gauge blocks are first wrung to what the standards call an auxiliary surface. We will call thesesurfaces platens. The platen can be made of any hard material, but are usually steel or glass. Anoptical flat is positioned above the gauge block, as in the flatness measurement, and the fringepatterns are observed. Figure 3.2 illustrates a typical fringe pattern. The angle between the referenceflat and gauge block is adjusted to orient the fringes across the width of the face as in Figure 3.2 oralong the length of the face. The reference flat is also adjusted to control the number of fringes,preferably 4 or 5 across, and 2 or 3 along. Four fringes in each direction are satisfactory for squareblocks.

Figure 3.2 Typical fringe patterns for measuring gauge block parallelism using theinterferometer method.

A parallelism error between the two faces is indicated by the slope of the gauge block fringes relativeto the platen fringes. Parallelism across the block width is illustrated in figure 3.2a where

Slope = (a/b) - (a'/b) = 0.8 - 0.3 = 0.5 fringe (3.1) Parallelism along the block length in figure 3.2b is

Slope = (a/b) + (a')/b = 0.8 + 0.3 = 1.1 fringe (3.2)

Note that the fringe fractions are subtracted for figure 3.2a and added for figure 3.2b. The reasonfor this is clear from looking at the patterns - the block fringe stays within the same two platen

b a

a

b

b

b

a'

a'

16

fringes in the first case and it extends into the next pair in the latter case. Conversion to length unitsis made with the value of λ/2 appropriate to the illumination.

Since a fringe represents points of equal elevation it is easy to visualize the blocks in figure 3.2 asbeing slightly wedge shaped.

This method depends on the wringing characteristics of the block. If the wringing is such that theplaten represents an extension of the lower surface of the block then the procedure is reliable. Thereare a number of problems that can cause this method to fail. If there is a burr on the block or platen,if there is particle of dust between the block and platen, or if the block is seriously warped, the entireface of the block may not wring down to the platen properly and a false measurement will result. For this reason usually a fused silica platen is used so that the wring can be examined by lookingthrough the back of the platen, as discussed in the section on flatness measurements. If the wringis good, the block-platen interface will be a fairly homogeneous gray color.

Gauge Block Comparator Technique

Electro-mechanical gauge block comparators with opposing measuring styli can be used to measureparallelism. A gauge block is inserted in the comparator, as shown in figure 3.3, after sufficienttemperature stabilization has occurred to insure that the block is not distorted by internal temperaturegradients. Variations in the block thickness from edge to edge in both directions are measured, thatis, across the width and along the length of the gauging face through the gauging point. Insulatedtongs are recommended for handling the blocks to minimize temperature effects during themeasuring procedure.

Figure 3.3. Basic geometry of measurements using a mechanical comparator.

Figures 3.4a and 3.4b show locations of points to be measured with the comparator on the twoprinciple styles of gauge blocks. The points designated a, b, c, and d are midway along the edgesand in from the edge about 0.5 mm ( 0.02 in) to allow for the normal rounding of the edges.

17

Figure 3.4 a and b. Location of gauging points on gauge blocks for both length (X)and parallelism (a,b,c,d) measurements.

A consistent procedure is recommended for making the measurements:

(1) Face the side of the block associated with point "a" toward the comparator measuringtips, push the block in until the upper tip contacts point "a", record meter reading andwithdraw the block.

(2) Rotate the block 180 º so the side associated with point "b" faces the measuring tips, push the block in until tip contacts point "b", record meter reading and

withdraw block.

(3) Rotate block 90 º so side associated with point "c" faces the tips and proceed as inprevious steps.

(4) Finally rotate block 180 ºand follow this procedure to measure at point "d".

The estimates of parallelism are then computed from the readings as follows:

Parallelism across width of block = a-b

Parallelism along length of block = c-d

The parallelism tolerances, as given in the GGG and ANSI standards, are shown in table 3.1.

x xa

a

bbc

c

d

d

18

Table 3.1 ANSI tolerances for parallelism in microinches

Size Grade .5 Grade 1 Grade 2 Grade 3(in)

<1 1 2 4 5 2 1 2 4 5

3 1 3 4 5 4 1 3 4 5 5-8 3 4 510-20 4 5 6

Referring back to the length tolerance table, you will see that the allowed parallelism and flatnesserrors are very substantial for blocks under 25 mm (or 1 in). For both interferometry and mechanicalcomparisons, if measurements are made with little attention to the true gauge point significant errorscan result when large parallelism errors exist.

3.3 Thermal Expansion

In most materials, a change in temperature causes a change in dimensions. This change depends onboth the size of the temperature change and the temperature at which the change occurs. Theequation describing this effect is

∆L/L = αL ∆T (3.3)

where L is the length, ∆L is the change in length of the object, ∆T is the temperature change and αLis the coefficient of thermal expansion(CTE).

3.3.1 Thermal Expansion of Gauge Block Materials

In the simplest case, where ∆T is small, αL can be considered a constant. In truth, αL depends on theabsolute temperature of the material. Figure 3.5 [18] shows the measured expansion coefficient ofgauge block steel. This diagram is typical of most metals, the thermal expansion rises withtemperature.

19

Figure 3.5. Variation of the thermal expansion coefficient of gauge block steel withtemperature.

As a numerical example, gauge block steel has an expansion coefficient of 11.5 x 10-6/ºC. Thismeans that a 100 mm gauge block will grow 11.5 x 10-6 times 100 mm, or 1.15 micrometer, whenits temperature is raised 1 ºC. This is a significant change in length, since even class 3 blocks areexpected to be within 0.2 µm of nominal. For long standards the temperature effects can bedramatic. Working backwards, to produce a 0.25 µm change in a 500 mm gauge block, atemperature change of only 43 millidegrees (0.043 ºC) is needed.

Despite the large thermal expansion coefficient, steel has always been the material of choice forgauge blocks. The reason for this is that most measuring and manufacturing machines are made ofsteel, and the thermal effects tend to cancel.

To see how this is true, suppose we wish to have a cube made in the shop, with a side length of100 mm. The first question to be answered is at what temperature should the length be 100 mm. As we have seen, the dimension of most objects depends on its temperature, and therefore adimension without a defined temperature is meaningless. For dimensional measurements thestandard temperature is 20 ºC (68 ºF). If we call for a 100 mm cube, what we want is a cube whichat 20 ºC will measure 100 mm on a side.

Suppose the shop floor is at 25 ºC and we have a perfect gauge block with zero thermal expansioncoefficient. If we make the cube so that each side is exactly the same length as the gauge block, whatlength is it? When the cube is taken into the metrology lab at 20 ºC, it will shrink 11.5 x 10-6 /ºC,which for our block is 5.75 µm, i.e., it will be 5.75 µm undersized.

20

Now suppose we had used a steel gauge block. When we brought the gauge block out onto the shopfloor it would have grown 5.75 µm. The cube, being made to the dimension of the gauge blockwould have been oversized by 5.75 µm. And finally, when the block and cube were brought into thegauge lab they would both shrink the same amount, 5.75 µm, and be exactly the length called for inthe specification.

What this points out is that the difference in thermal expansion between the workpiece and the gaugeis the important parameter. Ideally, when making brass or aluminum parts, brass or aluminumgauges would be used. This is impractical for a number of reasons, not the least of which is that itis nearly impossible to make gauge blocks out of soft materials, and once made the surface wouldbe so easily damaged that its working life would be on the order of days. Another reason is that mostmachined parts are made from steel. This was particularly true in the first half of the century whengauge blocks were invented because aluminum and plastic were still undeveloped technologies.

Finally, the steel gauge block can be used to gauge any material if corrections are made for thedifferential thermal expansion of the two materials involved. If a steel gauge block is used to gaugea 100 mm aluminum part at 25 ºC, a correction factor must be used. Since the expansion coefficientof aluminum is about twice that of steel, when the part is brought to 20 ºC it will shrink twice asmuch as the steel. Thus the aluminum block must be made oversized by the amount

∆L = (αLaluminum - αL

steel) x L x ∆T (3.4)

= (24 - 11.5) x 10-6 x 100 mm x 5ºC

= 6.25 µm

So if we make the cube 6.25 µm larger than the steel gauge block it will be exactly 100 mm whenbrought to standard conditions (20 ºC).

There are a few mixtures of materials, alloys such as invar and crystalline/glass mixtures such asZerodur, which have small thermal expansion coefficients. These materials are a combination of twocomponents, one which expands with increasing temperature, and one which shrinks with increasingtemperature. A mixture is made for which the expansion of one component matches the shrinkageof the other. Since the two materials never have truly opposite temperature dependencies, thematching of expansion and shrinkage can be made at only one temperature. It is important toremember that these materials are designed for one temperature, usually 20 ºC, and the thermalexpansion coefficient can be much different even a few degrees away. Examples of such materials,super-invar [19], Zerodur and Cervit [20], are shown in figure 3.6.

21

Figure 3.6. Variation of the thermal expansion coefficient for selected lowexpansion materials with temperature.

The thermal expansion coefficients, at 20 ºC, of commonly used materials in dimensional metrologyare shown in table 3.2.

Table 3.2

Material Thermal Expansion Coefficient

(10-6/ºC)

Aluminum 24Free Cutting Brass 20.5Steel Gauge Block ( <25mm ) 11.5Steel Gauge Block ( 500 mm ) 10.6Ceramic Gauge Block (zirconia) 9.2Chrome Carbide 8.4Granite 6.3 Oak (across grain) 5.4Oak (along grain) 4.9Tungsten Carbide 4.5Invar 1.2 Fused Silica 0.55Zerodur 0.05

22

To give a more intuitive feel for these numbers, figure 3.7 shows a bar graph of the relative changesin length of 100 mm samples of various materials when taken from 20 ºC to 25 ºC (68 ºF to 77 ºC).

Figure 3.7 Thermal Expansion of 100 mm blocks of various materials from 20 ºCto 25 ºC

3.3.2 Thermal Expansion Uncertainty

There are two sources of uncertainty in measurements due to the thermal expansion of gauges. Theseare apparent in thermal expansion equation, 3.2, where we can see that the length of the blockdepends on our knowledge of both the temperature and the thermal expansion coefficient of thegauge. For most measurement systems the uncertainty in the thermometer calibrations is known,either from the manufacturers specifications or the known variations from previous calibrations. Forsimple industrial thermocouple or thermistor based systems this uncertainty is a few tenths of adegree. For gauging at the sub-micrometer level this is generally insufficient and more sophisticatedthermometry is needed.

The uncertainty in the expansion coefficient of the gauge or workpiece is more difficult to estimate. Most steel gauge blocks under 100 mm are within a five tenths of 11.5 x 10-6/ºC, although there issome variation from manufacturer to manufacturer, and even from batch to batch from the samemanufacturer. For long blocks, over 100 mm, the situation is more complicated. Steel gauge blockshave the gauging surfaces hardened during manufacturing so that the surfaces can be properlylapped. This hardening process affects only the 30 to 60 mm of the block near the surfaces. Forblocks under 100 mm this is the entire block, and there is no problem. For longer blocks, there isa variable amount of the block in the center which is partially hardened or unhardened. Hardened

micrometers0 5 10 15 20 25

Zerodur

Fused Silica

InvarTungsten Carbide

Oak

Granite

Chrome CarbideSteel GB (500mm)

Steel GB (25mm)

Aluminum

Brass

23

steel has a higher thermal expansion coefficient than unhardened steel, which means that the longerthe block the greater is the unhardened portion and the lower is the coefficient. The measuredexpansion coefficients of the NIST long gauge blocks are shown in table 3.3.

Table 3.3. Thermal Expansion Coefficients of NIST Master Steel Gauge Blocks (10-6/ºC )

Size Set 1 Set 2(in.)

5 11.41 11.276 11.33 11.257 11.06 11.358 11.22 10.9210 10.84 10.6412 10.71 10.6416 10.80 10.5820 10.64 10.77

Table 3.3 shows that as the blocks get longer, the thermal expansion coefficient becomessystematically smaller. It also shows that the differences between blocks of the same size can be aslarge as a few percent. Because of these variations, it is important to use long length standards asnear to 20 ºC as possible to eliminate uncertainties due to the variation in the expansion coefficient.

As an example, suppose we have a 500 mm gauge block, a thermometer with an uncertainty of0.1ºC, and the thermal expansion coefficient is known to ± 0.3 x10-6. The uncertainties when thethermometer reads 20 and 25 degrees are

∆L = αLL x δ(∆T) + δ(αL)L x ∆T (3.4)

where ∆T is the temperature difference (T-20), and δ() denotes the uncertainty of the quantity withinthe parentheses.

At 25 ºC:

∆L = (11.5x10-6) x 500 x 0.1 + 0.3x10-6 x 500 x 5 (3.5) = 0.58 µm + .75 µm = 1.33 µm

24

At 20 ºC: when the thermometer reads 20 ºC, the worst case error is 0.1 ºC

∆L = (11.5x10-6) x 500 x 0.1 + 0.3x10-6 x 500 x .1 (3.6) = 0.58 µm + 0.02 µm = 0.60 µm

This points out the general need to keep dimensional metrology labs at, or very near 20 ºC.

3.4 Elastic Properties

When a force is exerted on any material, the material deforms. For steel and other gauge blockmaterials this effect is small, but not completely negligible. There are two-dimensional effects dueto the elastic properties of gauge blocks. The first, and least important, is the compression of blocksunder their own weight. When a block is supported horizontally, the force on each point is theweight of the steel above it, and the steel is slightly compressed. The compression is, however, notin the direction of the gauging dimension of the block and the effect is negligible. If the block is setupright, the force is now in the direction of the gauging surfaces, and for very long blocks the weightof the block can become significant. Solved analytically, the change in length of a block is foundto be

∆L = ρgL2/2E (3.7)

Where∆L = length of shorteningρ = density of materialg = acceleration of gravityL = total length of blockE = Young's modulus for material

For steel gauge blocks, the shrinkage is

∆L = (7.8x103(kg/m3) x 9.8 m/s2 x L2)/(2 x 210x109 N/m2) (3.8)

= 0.18x10-6 x L2 in meters. For a 500 mm gauge block the correction is 45 nm (1.8 µin). The corrections from this formula aremade at NIST on demand, but are negligible for blocks less than 300 mm (12 in).

When using long gauge blocks supported horizontally, some care is needed to assure that the blockbends properly. Since the sides of the gauge block are not precision surfaces, no matter how flat thesurface where it is placed it will touch only at a few points, therefore bending, and in generalproducing some small angle between the two gauging faces. The proper way to support the block

25

so that the two end faces are parallel, and thereby produce an unambiguous length, is shown in figure3.8. This assumes, however, that the gauging faces are parallel when the block is vertical.

Figure 3.8. Long gauge block supported at its Airy points.

When a block of length L is supported at two positions, 0.577L apart, the end faces will be parallel. These positions are called the Airy points.

3.4.1 Contact Deformation in Mechanical Comparisons Nearly all gauge block length comparisons or length measurements of objects with gauge blocks aremade with contact type comparators where a probe tip contacts a surface under an applied force. Contact between a spherical tip and a plane surface results in local deformation of small butsignificant magnitude. If the gauge blocks or objects being compared are made of the same material,the measured length difference between them will be correct, since the deformation in each case willbe the same. If the materials are different, the length difference will be incorrect by the differencein the amount of deformation for the materials. In such cases, a deformation correction may beapplied if its magnitude is significant to the measurement.

Total deformation (probe plus object) is a function of the geometry and elastic properties of the twocontacting surfaces, and contact force. Hertz [21] developed formulas for total uniaxial deformationbased on the theory of elasticity and by assuming that the bodies are isotropic, that there is notangential force at contact, and that the elastic limit is not exceeded in the contact area. Manyexperimenters [22, 23] have verified the reliability of the Hertzian formulas. The formulas givenbelow are from a CSIRO (Australian metrology laboratory) publication that contains formulas fora number of combinations of geometric contact between planes, cylinders and spheres [24]. The

26

gauge block deformations have been tested against other calculations and agree to a few nanometers.

For a spherical probe tip and a flat object surface the uniaxial deformation of the probe and surfacetogether is given by:

α = 3π2/3/(2P2/3(V1+V2)D1-1/3) (3.9)

WhereV1 = (1 - σ1

2)/πE1σ1 = Poisson ratio of sphere E1 = elastic modulus of sphere

V2 = (1 - σ22)/πE2

σ2 = Poisson ratio of block E2 = elastic modulus of block

P = force D = diameter of sphere

Some example deformations for a 6 mm diameter diamond stylus, and common gauge blockmaterials, at various pressures are given in table 3.4.

Table 3.4 Deformations at interface in micrometers (µin in parenthesis)

Material Force in Newtons 0.25 0.50 0.75

Fused Silica 0.13 (5.2) 0.21 (8.3) 0.28(11.2)Steel 0.07 (2.7) 0.11 (4.4) 0.14 (5.7)Chrome Carbide 0.06 (2.2) 0.12 (3.4) 0.12 (4.6)Tungsten Carbide 0.04 (1.6) 0.06 (2.5) 0.08 (3.2)

The gauge block comparators at NIST use 6 mm diameter tips , with forces of 0.25 N on the bottomprobe and 0.75 N on the top probe. The trade-offs involved in tip radius selection are:

1. The larger the probe radius the smaller the penetration, thus the correction will be smaller and lessdependent on the exact geometry of the tip radius.

2. The smaller the tip radius the greater its ability to push foreign matter, such as traces of oil, watervapor, or dust out of the way when the block is dragged between the tips.

27

There are comparator tips used by some laboratories that have radii as large as 10 to 20 mm. Theselarge radius tips have small penetrations, although generally more than 20 nm, and thus somecorrection still must be made.

3.4.2 Measurement of Probe Force and Tip Radius

Reliability of computed deformation values depends on careful measurement of probe force and,especially, of probe tip radius. Probe force is easily measured with a force gauge (or a double panbalance and a set of weights) reading the force when the probe indicator meter is at mid-scale on thehighest magnification range.

Probe tip inspection and radius measurement are critical. If the tip geometry is flawed in any wayit will not follow the Hertz predictions. Tips having cracks, flat spots, chips, or ellipticity should bereplaced and regular tip inspection must be made to insure reliability.

An interference microscope employing multiple beam interferometry is the inspection andmeasurement method used at NIST [25]. In this instrument an optical flat is brought close to the tip,normal to the probe axis, and monochromatic light produces a Newton Ring fringe pattern which ismagnified though the microscope lens system. The multiple beam aspect of this instrument isproduced by special optical components and results in very sharp interference fringes which revealfine details in the topography of the tip.

Figure 3.9 a, b, and c. Examples of microinterferograms of diamond stylus tips.

Figure 3.9 are multiple beam interference micrographs of diamond probe tips. The pictures havebeen skeleltonized so that the important features are clear. The micrograph is a "contour map" ofthe tip so that all points on a given ring are equidistant from the reference optical flat. This can beexpressed mathematically as

Nλ = 2t (3.10)

28

where N is the number (order) of the fringe, counting from the center, λ is the wavelength of the lightand t is the distance of the ring from the optical flat. The zero order fringe at the center is where theprobe tip is in light contact with the optical flat and t is nearly zero. This relationship is used tocalculate the tip radius.

Tip condition is readily observed from the micrograph. In figure 3.9a a crack is seen in the tip aswell as some ellipticity. The crack produces sharp breaks and lateral displacement of the fringesalong the crack. Figure 3.9b shows a sharp edge at the tip center. An acceptable tip is shown infigure 3.9c. It is very difficult to produce a perfectly spherical surface on diamond because thediamond hardness is not isotropic, i.e., there are "hard" directions and "soft" directions, and materialtends to be lapped preferentially from the "soft" directions.

Radius measurement of a good tip is relatively simple. Diameters of the first five rings are measuredfrom the photograph along axes A and B, as in figure 3.9c. If there is a small amount of ellipticity,A and B are selected as the major and minor axes. Then

dnrd = -------- (3.11)

2M

where rd is the actual radius of the nth Newton ring, dn is the ring average diameter (of A or B)measured on the micrograph, and M is the microscope magnification. Substituting the ring diametermeasurements in the equation will result in 5 radii, r1 through r5 and from these a radius of curvaturebetween consecutive rings is calculated:

ri+12 - ri2 Ri = ------------- (3.12)

λfor i = 1 to 4. The average of these four values is used as the tip radius.

The preceding measured and calculated values will also serve to evaluate tip sphericity. If theaverage difference between the five A and B ring diameters exceeds 10 percent of the average ringdiameter, there is significant lack of sphericity in the tip. Also, if the total spread among the fourtip radius values exceeds 10 percent of the average R there is significant lack of sphericity. Thesetests check sphericity around two axes so it is important that a tip meet both requirements or it willnot follow the Hertz prediction.

Our current laboratory practice uses only like materials as master blocks for comparisons. By havingone set of steel masters and one set of chrome carbide masters, the only blocks which havedeformation corrections are tungsten carbide. We have too small a customer base in this materialto justify the expense of a third master set. This practice makes the shape of the comparator tipsunimportant, except for cracks or other abnormalities which would scratch the blocks.

29

3.5 Stability

No material is completely stable. Due to processes at the atomic level all materials tend to shrinkor grow with time. The size and direction of dimensional change are dependent on the fabricationprocesses, both the bulk material processing as well as the finishing processing. During the 1950'sthe gauge block manufacturers and an interdisciplinary group of metallurgists, metrologists andstatisticians from NIST (NBS at the time) did extensive studies of the properties of steel gaugeblocks to optimize their dimensional stability. Blocks made since that era are remarkably stablecompared to their predecessors, but not perfect. Nearly all NIST master gauge blocks are very stable. Two typical examples are shown in figure 3.10. Because most blocks are so stable, we demand ameasurement history of at least 5 years before accepting a non-zero slope as real.

Figure 3.10. Examples of the dimensional stability of NIST master gauge blocks.

A histogram of the growth rates of our master blocks, which have over 15 years of measurementhistory, is shown in figure 3.11. Note that the rate of change/unit of length is the pertinentparameter, because most materials exhibit a constant growth rate per unit of length.

30

Figure 3.11. Histogram of the growth rates of NIST master blocks.There have been studies of dimensional stability for other materials, and a table of typical values [26-31] is given in table 3.5. In general, alloys and glassy materials are less stable than compositematerials. For the average user of gauge blocks the dimensional changes in the time betweencalibration is negligible.

Table 3.5

Material Stability (1 part in 106/yr)

Zerodur -0.7 to -0.03Corning 7971 ULE -0.14, 0.07, 0.06Corning 7940(fused Silica) -0.18, -0.18Cer-Vit C-1010 0.54, -0.18 LR-35 Invar 2.05Super Invar 0.052100 Gauge Block Steel 0.01Brass -1.2

Steel, chrome carbide and tungsten carbide have all proven to be very stable materials for gaugeblocks. The newer ceramic materials may also be stable, but no studies have been made of eitherthe material or gauge blocks made from it. Until studies are available, a shorter calibration cycle forthe first few years should be considered to assure that the blocks are suitably stable.

0

5

10

15

20

25

30

35

40

Rate of Change (ppm/yr)

Num

ber o

f Blo

cks

31

4. Measurement Assurance Program

4.1 Introduction

One of the primary problems in metrology is to estimate the uncertainty of a measurement. Traditional methods, beginning with simple estimates by the metrologist based on experiencegradually developed into the more formal error budget method. The error budget method, whileadequate for many purposes, is slowly being replaced by more objective methods based onstatistical process control ideas. In this chapter we will trace the development of these methodsand derive both an error budget for gauge block calibrations and present the measurementassurance (MAP) method currently used at NIST.

4.2 A comparison: Traditional Metrology versus Measurement Assurance Programs

4.2.1 Tradition

Each measurement, in traditional metrology, was in essence a "work of art." The result wasaccepted mainly on the basis of the method used and the reputation of the person making themeasurement. This view is still prevalent in calibrations, although the trend to statistical processcontrol in industry and the evolution of standards towards demanding supportable uncertaintystatements will eventually end the practice.

There are circumstances for which the "work of art" paradigm is appropriate or even inevitableas, for example, in the sciences where most experiments are seldom repeated because of the timeor cost involved, or for one of a kind calibrations . Since there is no repetition with which toestimate the repeatability of the measurement the scientist or metrologist is reduced to making alist of possible error sources and estimating the magnitude of these errors. This list, called theerror budget, is often the only way to derive an uncertainty for an experiment.

Making an accurate error budget is a difficult and time consuming task. As a simple example,one facet of the error budget is the repeatability of a gauge block comparison. A simpleestimation method is to measure the same block repeatedly 20 or more times. The standarddeviation from the mean then might be taken as the short term repeatability of the measurement. To get a more realistic estimate that takes into consideration the effects of the operator andequipment, the metrologist makes a GR&R (gauge repeatability and reproducibility) study. Inthis test several blocks of different length are calibrated using the normal calibration procedure. The calibrations are repeated a number of times by different operators using differentcomparators. This data is then analyzed to produce a measure of the variability for themeasurement process. Even a study as large as this will not detect the effects of long termvariability from sources such as drift in the comparator or thermometer calibrations. Unfortunately, these auxiliary experiments are seldom done in practice and estimates based onthe experimenter's experience are substituted for experimentally verified values. These estimatesoften reflect the experimenter's optimism more than reality.

32

The dividing line between the era of traditional metrology and the era of measurement assuranceprograms is clearly associated with the development of computers. Even with simplecomparison procedures, the traditionalist had to make long, detailed hand computations. A largepercentage of calibration time was spent on mathematical procedures, checking and doublechecking hand computations. The calibration of a meter bar could generate over a hundred pagesof data and calculations. Theoretical work revolved around finding more efficient methods,efficient in the sense of fewer measurements and simpler calculations.

4.2.2 Process Control: A Paradigm Shift

Measurement assurance has now become an accepted part of the national standards system. According to the ANSI/ASQC Standard M-1, "American National Standard for CalibrationSystems" the definition of measurement assurance is as follows:

2.12 Measurement Assurance Method

A method to determine calibration or measurement uncertainty based onsystematic observations of achieved results. Calibration uncertainty limitsresulting from the application of Measurement Assurance Methods are consideredestimates based upon objective verification of uncertainty components.

Admittedly, this definition does not convey much information to readers who do not alreadyunderstand measurement. To explain what the definition means, we will examine the more basicconcept of process control and show how it relates to measurement and calibration. Measurement assurance methods can be thought of as an extension of process control.

The following is an example of control of a process using a model from the chemical industry. In a chemical plant the raw materials are put into what is essentially a black box. At themolecular level, where reactions occur, the action is on a scale so small and so fast that no humanintervention is possible. What can be measured are various bulk properties of the chemicals; thetemperature, pressure, fluid flow rate, etc. The basic model of process control is that once thefactory output is suitable, if the measurable quantities (process control variables) are keptconstant the output will remain constant. Thus the reactions, which are not observable, aremonitored and characterized by these observable control variables.

Choosing control variables is the most important part of the control scheme. A thermometerplaced on a pipeline or vat may or may not be a useful measure of the process inside. Somechemical reactions occur over a very large range of temperatures and the measurement andcontrol of the temperature might have virtually no effect on the process. In another part of theplant the temperature may be vitally important. The art and science of choosing the processcontrol measurements is a large part of the chemical engineering profession.

Defining the measurement process as a production process evolved in the early 1960's [33], atime coincident with the introduction of general purpose computers on a commercial basis. Both

33

the philosophy and scope of measurement assurance programs are a direct result of being able tostore and recall large amounts of data, and to analyze and format the results in many differentways. In this model, the calibration procedure is a process with the numbers as the processoutput. In our case, the length of a gauge block is the output.

The idea of a measurement assurance program is to control and statistically characterize ameasurement process so that a rational judgement about the accuracy the process can be made. Once we accept the model of measurement as a production process we must examine the processfor possible control variables, i.e., measurable quantities other than the gauge block length whichcharacterize the process.

For example, suppose we compare one gauge block with one "master" block. If one comparisonis made, the process will establish a number for the length of the block. An obvious extension ofthis process is to compare the blocks twice, the second measurement providing a measure of therepeatability of the process. Averaging the two measurements also will give a more accurateanswer for two reasons. First, statistics assures us that the mean of several measurements has ahigher probability of being the accurate than any single measurement. Secondly, if the twomeasurements differ greatly we can repeat the measurement and discard the outlyingmeasurement as containing a blunder. Thus the repeatability can be used as a process controlvariable.

The next extension of the method is to make more measurements. It takes at least fourmeasurements for statistical measures of variability to make much sense. At this point thestandard deviation becomes a possible statistical process control (SPC) parameter. The standarddeviation of each measurement is a measure of the short term repeatability of the process whichcan be recorded and compared to the repeatability of previous measurements. When this shortterm repeatability is much higher than its historical value we can suspect that there is a problemwith the measurement, i.e., the process is out of control. Since the test is statistical in nature wecan assign a confidence level to the decision that the process is in or out of control. An analysisof the economic consequences of accepting bad or rejecting good calibrations can be used tomake rational decisions about the control limits and confidence levels appropriate to thedecisions.

Another extension is to have two master blocks and compare the unknown to both of them. Sucha procedure can also give information about the process, namely the observed difference betweenthe two masters. As in the previous example, this difference can be recorded and compared tothe differences from previous calibrations. Unlike the previous case this process control variablemeasures long term changes in the process. If one of the blocks is unstable and grows, over ayear the difference between the masters will change and an examination of this control parameterwill detect it. Short term process repeatability is not affected by the length changes of the blocksbecause the blocks do not change by detectable amounts during the few minutes of a comparison.

Thus, comparing a gauge block with two masters according to a measurement design provides,not only a value for the block, but also an estimate of short term process repeatability, and in

34

time, an estimate of long term process variability and a check on the constancy of the masters. All of this is inexpensive only if the storage and manipulation of data is easy to perform and fast,tasks for which computers are ideal.

The complexity of a measurement assurance program depends upon the purpose a particularmeasurement is to serve. NIST calibration services generally aim at providing the highestpractical level of accuracy. At lower accuracy levels the developer of a measurement systemmust make choices, balancing the level of accuracy needed with costs. A number of differentapproaches to measurement assurance system design are discussed by Croarkin [34].

4.2.3 Measurement Assurance: Building a Measurement Process Model

Assigning a length value to a gauge block and determining the uncertainty of that value is not asimple task. As discussed above, the short and long term variability of the measurement processare easily measured by redundant measurements and a control measurement (the differencebetween two masters) in each calibration. This is not enough to assure measurement accuracy. SPC only attempts to guarantee that the process today is the same as the process in the past. Ifthe process begins flawed, i.e., gives the wrong answer, process control will only guarantee thatthe process continues to give wrong answers. Obviously we need to ascertain the sources of errorother than the variability measured by the SPC scheme.

Many auxiliary parameters are used as corrections to the output of the measurement procedure. These corrections must be measured separately to provide a complete physical model for themeasurement procedure. Examples of these parameters are the phase change of light onreflection from the gauge block during interferometric measurements, the penetration depth ofthe gauge block stylus into the block during mechanical comparisons, and thermal expansioncoefficients of the blocks during all measurements. These auxiliary parameters are sources ofsystematic error.

To assess systematic errors, studies must be made of the effects of factors not subject to control,such as the uncertainty of thermometer and barometer calibrations, or variations in thedeformability of gauge blocks. Combining these systematic errors with the known random error,one arrives at a realistic statement of the accuracy of the calibration.

In summary, the MAP approach enables us to clearly establish limitations for a particularmeasurement method. It combines process control with detailed process modeling. The processcontrol component provides a means to monitor various measurement parameters throughout thesystem and provides estimates of the random error. The calibration model allows us to obtainreasonable estimates of the systematic uncertainties that are not sampled by the statistics derivedfrom an analysis of the process control parameters. This leads to a detailed understanding of themeasurement system and an objective estimate of the measurement uncertainty. It also creates areliable road map for making process changes that will increase calibration accuracy.

35

4.3 Determining Uncertainty

4.3.1 Stability

All length measurement processes are, directly or indirectly, comparative operations. Even thesimplest concept of such a process assumes that the basic measurement unit (in our case themeter) is constant and the object, procedures, equipment, etc., are stable.

As an example, before 1959 the English inch was defined in terms of an official yard bar and theAmerican inch was defined in terms of the meter. By 1959 the English inch had shrunkcompared to the American inch because the yard bar was not stable and was shrinking withrespect to the meter. This problem was solved by adopting the meter as the primary unit oflength and defining the international inch as exactly 25.4 mm.

Similarly, the property to be measured must be predictable. If a measurement process detects adifference between two things, it is expected that repeated measures of that difference shouldagree reasonably well. In the absence of severe external influence, one does not expect things tochange rapidly.

There is a difference between stability and predictability as used above. Repeated measurementsover time can exhibit a random-like variability about a constant value, or about a time dependentvalue. In either case, if the results are not erratic (with no unexpected large changes), the processis considered to be predictable. Gauge blocks that are changing length at a constant rate can beused because they have a predictable length at any given time. Stability means that thecoefficients of time dependent terms are essentially zero. Stability is desirable for certain uses,but it is not a necessary restriction on the ability to make good measurements.

4.3.2 Uncertainty

A measurement process is continually affected by perturbations from a variety of sources. Therandom-like variability of repeated measurements is a result of these perturbations. Randomvariability implies a probability distribution with a range of variability that is not likely to exceedtest limits. Generally a normal distribution is assumed.

Traditionally the second type of error is called systematic error, and includes uncertainties thatcome from constants that are in error and discrepancies in operational techniques. Thesystematic error, expressed as a single number, is an estimate of the offset of the measurementresult from the true value. For example, for secondary laboratories using NIST values for theirmaster blocks the uncertainty reported by NIST is a systematic error. No matter how often themaster blocks are used the offset between the NIST values and the true lengths of the blocksremains the same. The random error and systematic error are combined to determine theuncertainty of a calibration.

A new internationally accepted method for classifying and combining errors has been developed

36

by the International Bureau of Weights and Measures (BIPM) and the International Organizationfor Standardization (ISO). This method seems more complicated than the traditional method,and is presented in detail in the ISO "Guide to the Expression of Uncertainty in Measurement"[35]. A short summary of the method and how it will be applied at NIST is given in reference[36]. Quoting ISO, the core concepts are:

1. The uncertainty in the result of a measurement generally consists of severalcomponents which may be grouped into two categories according to the way in whichtheir numerical value is estimated:

A - those which are evaluated by statistical methods,B - those which are evaluated by other means.

There is not always a simple correspondence between the classification into categories Aor B and the previously used classification into "random" and "systematic" uncertainties. The term "systematic uncertainty" can be misleading and should be avoided.

Any detailed report of the uncertainty should consist of a complete list of the components,specifying for each the method used to obtain its numerical value.

2. The components in category A are characterized by estimated variances, si2, (or the

estimated "standard deviations" si) and the number of degrees of freedom, vi . Whereappropriate, the estimated covariances should be given.

3. The components in category B should be characterized by quantities uj2, which may be

considered as approximations to the corresponding variances, the existence of which isassumed. The quantities uj

2 may be treated like variances and the quantities uj likestandard deviations. Where appropriate, the covariances should be treated in a similarway.

4. The combined uncertainty should be characterized by the numerical value obtained byapplying the usual method for the combination of variances. The combined uncertaintyand its components should be expressed in the form of "standard deviations."

5. If, for particular applications, it is necessary to multiply the combined uncertainty by afactor to obtain an overall uncertainty, the multiplying factor used must always be stated.

The random errors, being determined statistically, are type A errors. The errors that made up thesystematic errors become the sources of type B uncertainty. The main difference is the methodfor estimating the type B uncertainty. The traditional systematic error was estimated by themetrologist with little or no guidance. The ISO guide gives specific criteria for the estimation oftype B uncertainty. NIST has adopted the new international method for reporting calibrationuncertainties. In addition to adopting the new method, NIST has adopted k=2 as the coveragefactor, rather than the more traditional 3σ, to align its reported uncertainty with the current

37

International practice.

4.3.3 Random Error

There are very few sources of truly random error. Some examples are 1/f and shot noise inelectronic components. These contributions to the uncertainty of most dimensionalmeasurements are very small. Most of the contributions to the random errors are systematiceffects from sources that are not under experimental control. Examples are floor and airvibrations, thermal drift, and operator variability. Many of these error sources are too costly tocontrol.

These effects can be grouped into two categories: long term effects in which the effect does notsignificantly change in the time required for a given sequence of measurements and short termeffects which vary significantly in the course of a single measurement or measurements madeover a short time interval. Typical long term effects are changes in length of the master blocks orslow changes in the comparator electronic magnification.

An example of smooth continuous short term effects are thermal drift of the environment duringthe measurement. Another category of short term effects are those which are instantaneous, orstep-like, in nature. A simple example is a dust particle left on a gauge block. Depending on theexact point of contact the gauge length may be shifted by the size of the particle. In many cases,"shocks" or vibrations can cause temporary shifts in the instrument output.

These short term variations make up the within-group variability usually expressed as a standarddeviation, σw. The within-group variability of the measurement process is the most familiarprocess parameter, and is generally called repeatability. It is easily demonstrated in a repeatedsequence of measurements of the same thing in a short time interval. Practically allmeasurements should be repeated several times. Within-group variability is generally limited bythe degree that certain types of perturbations are controlled and by factors such as operator skills,instrument quality, and attention to detailed procedure. Sources contributing to within variabilityare not easily identified and process improvement in terms of reducing σw is obtained morefrequently by trial and error than by design.

Process accuracy is often judged on the basis of within-group variability. Such a judgment,however, is usually erroneous. A more reliable assessment is the total variability that is thevariability of a long sequence of measurements. Repeating a given measurement over a timeinterval sufficiently long to reflect the influence of all possible perturbations establishes a totalprocess standard deviation, σt.

The primary way to improve total variability is to modify the process by reducing the variabilityassociated with all identifiable perturbations. This can be very tedious and expensive, but isnecessary to produce the highest quality measurements.

38

Another straightforward way to reduce variability is to place severe restrictions on thecharacteristics of objects being measured and on the measurement environment. Reducing thecomplexity of the object reduces the possible error sources. Thus traditional metrology focusedon simple geometric forms, spheres, cylinders and blocks as the mainstays of length standards. By making all measurements at or near standard conditions, 20 ºC for example, many other errorsare reduced. Restricting the number and severity of possible errors brings the uncertaintyestimate closer to the truth.

4.3.4 Systematic Error and Type B Uncertainty

The sources of systematic error and type B uncertainty are the same. The only major differenceis in how the uncertainties are estimated. The estimates for systematic error which have beenpart of the NIST error budget for gauge blocks have been traditionally obtained by methodswhich are identical or closely related to the methods recommended for type B uncertainties.

There are several classes of type B uncertainty. Perhaps the most familiar is instrument scalezero error (offset). Offset systematic errors are not present in comparative measurementsprovided that the instrument response is reasonably linear over the range of difference that mustbe measured. A second class of type B uncertainty comes from supplemental measurements. Barometric pressure, temperature, and humidity are examples of interferometric supplementalmeasurements used to calculate the refractive index of air, and subsequently the wavelength oflight at ambient conditions. Other examples are the phase shift of light on reflection from theblock and platen, and the stylus penetration in mechanical comparisons. Each supplementalmeasurement is a separate process with both random variability and systematic effects. Therandom variability of the supplemental measurements is reflected in the total process variability. Type B Uncertainty associated with supplemental data must be carefully documented andevaluated.

One practical action is to "randomize" the type B uncertainty by using different instruments,operators, environments and other factors. Thus, variation from these sources becomes part ofthe type A uncertainty. This is done when we derive the length of master blocks from theirinterferometric history. At NIST, we measure blocks using both the Hilger1 and Zeiss1

interferometers and recalibrate the barometer, thermometers, and humidity meters before startinga round of calibrations. As a measurement history is produced, the effects of these factors arerandomized. When a block is replaced this "randomization" of various effects is lost and can beregained only after a number of calibrations.

Generally, replacement blocks are not put into service until they have been calibratedinterferometrically a number of times over three or more years. Since one calibration consists of3 or more wrings, the minimum history consists of nine or more measurements using three

1 Certain commercial instruments are identified in this paper in order to adequately specify our

measurement process. Such identification does not imply recommendation or endorsement by NIST.

39

different calibrations of the environmental sensors. While this is merely a rule of thumb,examination of the first 10 calibrations of blocks with long histories appear to assume values andvariances which are consistent with their long term values.

The traditional procedure for blocks that do not have long calibration histories is to evaluate thetype B uncertainty by direct experiment and make appropriate corrections to the data. In manycases the corrections are small and the uncertainty of each correction is much smaller than theprocess standard deviation. For example, in correcting for thermal expansion, the errorintroduced by a temperature error of 0.1 ºC for a steel gauge block is about 1 part in 106. This isa significant error for most calibration labs and a correction must be applied. The uncertainty inthis correction is due to the thermometer calibration and the uncertainty in the thermal expansioncoefficient of the steel. If the measurements are made near 20 ºC the uncertainty in the thermalcoefficient (about 10%) leads to an uncertainty in the correction of 1 part in 107 or less; anegligible uncertainty for many labs.

Unfortunately, in many cases the auxiliary experiments are difficult and time consuming. As anexample, the phase correction applied to interferometric measurements is generally about 25 nm(1 µin). The 2σ random error associated with measuring the phase of a block is about 50 nm(2µin). The standard deviation of the mean of n measurements is n-1/2 times the standarddeviation of the original measurements. A large number of measurements are needed todetermine an effect which is very small compared to the standard deviation of the measurement. For the uncertainty to be 10% of the correction, as in the thermal expansion coefficient example,n must satisfy the relation:

(4.1)

Since this leads to n=400, it is obvious this measurement cannot be made on every customerblock. Phase measurements made for a random sample of blocks from each manufacturer haveshown that the phase is nearly the same for blocks made (1) of the same material, (2) by the samelapping procedure and (3) by the same manufacturer. At NIST, we have made supplementarymeasurements under varying conditions and calculated compensation factors for eachmanufacturer (see appendix C). A corollary is that if you measure your master blocks byinterferometry, it helps to use blocks from only one or two manufacturers.

Our next concern is whether the corrections derived from these supplementary measurements areadequate. A collection of values from repeated measurements should be tested for correlationwith each of the supplementary measurements, and their various combinations. If correlation isindicated then the corrections that are being applied are not adequate. For example, if acorrelation exists between the length of gauge blocks previously corrected for temperature andthe observed temperature of the block, then either the temperature is being measured at aninappropriate location or the correction scheme does not describe the effect that is actuallyoccurring. Corrective action is necessary.

)nanometers(inn

50<25of10%

40

Low correlation does not necessarily indicate that no residual type B uncertainty are present, butonly that the combined type B uncertainty are not large relative to the total standard deviation ofthe process. There may be long term systematic error effects from sources not associated withthe current supplemental measurements. It is relatively easy to demonstrate the presence orabsence of such effects, but it may be difficult to reduce them. If both the short and long termstandard deviations of the process are available, they are expected to be nearly equal if there areno unresolved sources of variation. Frequently such is not the case. In the cases where the longterm SD is distinctly larger than the short term SD there are uncorrected error sources at work.

The use of comparison designs, described in chapter 5, facilitates this type of analysis. Thewithin group variability, σw, is computed for each measurement sequence. Each measurementsequence includes a check standard that is measured in each calibration. The total standarddeviation, σt is computed for the collection of values for the check standard. The inequalityσt>>σw is taken as evidence of a long term systematic effect, perhaps as yet unidentified. For theNIST system a very large number of measurements have been made over the years and we havefound, as shown in table 4.1, that except for long blocks there appear to be no uncorrected longterm error sources. The differences for long blocks are probably due to inadequate thermalpreparation or operator effects caused by the difficult nature of the measurement.

Table 4.1

Comparison of Short Term and Long Term Standard Deviationsfor Mechanical Gauge Block Comparisons (in nanometers)

Group σshort term σ σlong term

14 1.00 mm to 1.09 mm 5 615 1.10 mm to 1.29 mm 5 516 1.30 mm to 1.49 mm 5 517 1.50 mm to 2.09 mm 5 518 2.10 mm to 2.29 mm 5 519 2.30 mm to 2.49 mm 5 520 2.50 mm to 10 mm 5 521 10.5 mm to 20 mm 5 722 20.5 mm to 50 mm 6 823 60 mm to 100 mm 8 1824 125 mm 9 19

150 mm 11 38 175 mm 10 22 200 mm 14 37 250 mm 11 48 300 mm 7 64 400 mm 11 58 500 mm 9 56

41

4.3.5 Error Budgets

An error budget is a systematic method for estimating the limit of error or uncertainty for ameasurement. In this method, one lists all known sources of error which might affect themeasurement result. In the traditional method, a theoretical analysis of error sources is made toprovide estimates of the expected variability term by term. These estimates are then combined toobtain an estimate of the total expected error bounds.

The items in an error budget can be classified according to the way they are most likely to affectthe uncertainty. One category can contain items related to type B uncertainty; another can relateto σw; and a third category can relate to σt. Since all sources of short term variability also affectthe long term variability, short term effects are sometimes subtracted to give an estimate of theeffects of only the sources of long term variability. This long term or "between groupvariability", σB, is simply calculated as (σt-σw). The example in table 4.2 shows a tentativedisposition for the mechanical comparison of gauge blocks.

Table 4.2

Error Source Category

Uncertainty of Master Block systematic

Stylus diameter, pressure and condition long term

Assigned Temperature Coefficients systematic

Temperature difference between blocks long term

Stability of Blocks long term

Environment: vibration, thermal changes, operator effects short term

Block flatness/ anvil design systematic

Block geometry: effect of flatnessand parallelism on measurement with gauging point variability short term

Comparator calibration long term

Comparator repeatability short term

42

Most of the error sources in table 4.2 can be monitored or evaluated by a judicious choice of acomparison design, to be used over a long enough time so that the perturbations can exert theirfull effect on the measurement process. Two exceptions are the uncertainties associated with theassigned starting values, or restraint values, and the instrument calibration.

Table 4.3 shows error budget items associated with gauge block measurement processes. Listed(1σ level) uncertainties are typical of the NIST measurement process for blocks under 25 mmlong. The block nominal length, L, is in meters.

Table 4.3

Resulting Length Error Sources Uncertainty in µm Note

Repeatability 0.004 + 0.12 L (1)

Contact Deformation 0.002 (2)

Value of Master Block 0.012 + 0.08 L

Temperature DifferenceBetween Blocks 0.17 L (3)

Uncertainty of ExpansionCoefficient 0.20 L (4)

Note 1. Repeatability is determined by analyzing the long term control data. The numberreported is the 1σt value for blocks under 25 mm.

Note 2. For steel and chrome carbide, blocks of the same material are used in the comparatorprocess and the only error is from the variation in the surface hardness from block to block. Thenumber quoted is an estimate. For other materials the estimate will be higher.

Note 3. A maximum temperature difference of 0.030 ºC is assumed. Using this as the limit of arectangular distribution yields a standard uncertainty of 0.17L.

Note 4. A maximum uncertainty of 1x10-6/ºC is assumed in the thermal expansion coefficient,and a temperature offset of 0.3 ºC is assumed. For steel and chrome carbide blocks we have nomajor differential expansion error since blocks of the same material are compared. For othermaterials the estimate will be higher.These errors can be combined in various ways, as discussed in the next section.

43

4.3.6 Combining Type A and Type B uncertainties

Type A and type B uncertainties are treated as equal, statistically determined and if uncorrelatedare combined in quadrature.

In keeping with current international practice, the coverage factor k is set equal to two (k=2) onNIST gauge block uncertainty statements.

Combining the uncertainty components in Table 4.3 according to these rules gives a standarduncertainty (coverage factor k=2)

U = Expanded Uncertainty = 0.026 + 0.32L (4.3)

Generally, secondary laboratories can treat the NIST stated uncertainty as an uncorrelated type Buncertainty. There are certain practices where the NIST uncertainty should be used with caution. For example, if a commercial laboratory has a gauge block measured at NIST every year for nineyears, it is good practice for the lab to use the mean of the nine measurements as the length of theblock rather than the last reported measurement. In estimating the uncertainty of the mean,however, some caution is needed. The nine NIST measurements contain type B uncertaintiesthat are highly correlated, and therefore cannot be handled in a routine manner. Since the phaseassigned to the NIST master blocks, the penetration, wringing film thickness and thermalexpansion differences between the NIST master and the customer block do not change over thenine calibrations, the uncertainties due to these factors are correlated. In addition, the length ofthe NIST master blocks are taken from a fit to all of its measurement history. Since most blockshave over 20 years of measurement history, the new master value obtained when a few newmeasurements are made every 3 years is highly correlated to the old value since the fit is heavilyweighted by the older data. To take the NIST uncertainty and divide it by 3 (square root of 9)assumes that all of the uncertainty sources are uncorrelated, which is not true.

Even if the standard deviation from the mean was calculated from the nine NIST calibrations, theresulting number would not be a complete estimate of the uncertainty since it would not includethe effects of the type B uncertainties mentioned above. The NIST staff should be consultedbefore attempting to estimate the uncertainty in averaged NIST data.

4.3.7 Combining Random and Systematic Errors

There is no universally accepted method of combining these two error estimates. The traditionalway to combine random errors is to sum the variances characteristic of each error source. Sincethe standard deviation is the square root of the variance, the practice is to sum the standarddeviations in quadrature. For example, if the measurement has two sources of random errorcharacterized by σ1 and σ2, the total random error is

) (4.2 B+Ak =ty UncertainExpanded = U 22

44

σtot = [σ12 + σ2

2]1/2. (4.4)

This formula reflects the assumption that the various random error sources are independent. Ifthere are a number of sources of variation and their effects are independent and random, theprobability of all of the effects being large and the same sign is very small. Thus the combinederror will be less than the sum of the maximum error possible from each source.

Traditionally, systematic errors are not random (they are constant for all calibrations) and may behighly correlated, so nothing can be assumed about their relative sizes or signs. It is logical thento simply add the systematic errors to the total random error:

Total Uncertainty (traditions) = 3σtot + Systematic Error. (4.5)

The use of 3σ as the total random error is fairly standard in the United States. This correspondsto the 99.7% confidence level, i.e., we claim that we are 99.7% confident that the answer iswithin the 3σ bounds. Many people, particularly in Europe, use the 95% confidence level (2σ)or the 99% confidence level (2.5σ) rather than the 99.7% level (3σ) used above. This is a smallmatter since any time an uncertainty is stated the confidence level should also be stated. Theuncertainty components in Table 4.3 are all systematic except the repeatability, and theuncertainty calculated according to the traditional method is

Total Uncertainty (traditional) = 3(0.004 + 0.12L) + 0.002 + 0.012 + 0.08L + 0.17L + 0.2L

Total Uncertainty (traditional) = 0.026 + 0.81L

4.4 The NIST Gauge Block Measurement Assurance Program

The NIST gauge block measurement assurance program (MAP) has three parts. The first is acontinual interferometric measurement of the master gauge blocks under process control suppliedby control charts on each individual gauge block. As a measurement history is built up for eachblock, the variability can be checked against both the previous history of the block and the typicalbehavior of other similar sized blocks. A least squares fit to the historical data for each blockprovides estimates of the current length of the block as well as the uncertainty in the assignedlength. When the master blocks are used in the mechanical comparison process for customerblocks, the interferometric uncertainty is combined in quadrature with the Intercomparisonrandom error. This combined uncertainty is reported as the uncertainty on the calibrationcertificate.

The second part of the MAP is the process control of the mechanical comparisons with customerblocks. Two master blocks are used in each calibration and the measured difference in lengthbetween the master blocks is recorded for every calibration. Also, the standard deviation derived

45

from the comparison data is recorded. These two sets of data provide process control parametersfor the comparison process as well as estimates of short and long term process variability.

The last test of the system is a comparison of each customer block length to its history, if oneexists. This comparison had been restricted to the last calibration of each customer set, but sinceearly 1988 every customer calibration is recorded in a database. Thus, in the future we cancompare the new calibration with all of the previous calibrations. Blocks with large changes arere-measured, and if the discrepancy repeats the customer is contacted to explore the problem inmore detail. When repeated measurements do not match the history, usually the block has beendamaged, or even replaced by a customer. The history of such a block is then purged from thedatabase.

4.4.1 Establishing Interferometric Master Values

The general practice in the past was to use the last interferometric calibration of a gauge block asits "best" length. When master blocks are calibrated repeatedly over a period of years thepractice of using a few data points is clearly not the optimum use of the information available. Ifthe blocks are stable, a suitable average of the history data will produce a mean length with amuch lower uncertainty than the last measurement value. Reviewing the calibration history since the late fifties indicates a reasonably stable length formost gauge blocks. More information on stability is given in section 3.5. The general patternsfor NIST master gauge blocks are a very small constant rate of change in length over the timecovered. In order to establish a "predicted value," a line of the form:

Lt = Lo + K1(t-to) (4.6)

is fit to the data. In this relation the correction to nominal length (Lt) at any time (t) is a functionof the correction (Lo) at an arbitrary time (to), the rate of change (K1), and the time interval (t-to).

When the calibration history is short (occasionally blocks are replaced because of wear) theremay not be enough data to make a reliable estimate of the rate of change K1. We use a fewsimple rules of thumb to handle such cases:

1. If the calibration history spans less than 3 years the rate of change is assumed tobe zero. Few blocks under a few inches have rates of change over 5 nm/year, andthus no appreciable error will accumulate.

2. A linear least squares fit is made to the history (over 3 years) data. If thecalculated slope of the line does not exceed three times the standard deviation ofthe slope, we do not consider the slope to be statistically significant. The data isthen refitted to a line with zero slope.

46

Figure 4.1a. Interferometric history of the 0.1 inch steel master block showingtypical stability of steel gauge blocks.

Figure 4.1b. Interferometric history of the 0.112 inch steel master block showing a predictablechange in length over time.

47

Figure 4.1a shows the history of the 0.1 inch block, and it is typical of most NIST master blocks. The block has over 20 years of history and the material is very stable. Figure, 4.1b, shows a blockwhich has enough history to be considered unstable; it appears to be shrinking. New blocks thatshow a non-negligible slope are re-measured on a faster cycle and if the slope continues for over 20measurements and/or 5 years the block is replaced.

The uncertainty of predicted values, shown in figure 4.1 by lines above and below the fitted line, isa function of the number of points in each collection, the degree of extrapolation beyond the timespan encompassed by these points, and the standard deviation of the fit.

The uncertainty of the predicted value is computed by the relation

(4.7)

where n = the number of points in the collection;

t = time/date of the prediction;

ta= average time/date ( the centroid of time span covered by the measurement history, thatis ta=Σti/n);

ti= time/date associated with each of the n values;

σ = process standard deviation (s.d. about the fitted line);

σt= s.d. of predicted value at time t.

In figure 4.2 the standard deviations of our steel inch master blocks are shown as a function ofnominal lengths. The graph shows that there are two types of contributions to the uncertainty, oneconstant and the other length dependent.

)t-t(

)t-(t+n13 = 3 2

ai

2a

t

48

Figure 4.2. Plot of standard deviations of interferometric length history of NISTmaster steel gauge blocks versus nominal lengths.

The contributions that are not length dependent are primarily variations in the wringing filmthickness and random errors in reading the fringe fractions. The major length dependent componentare the refractive index and thermal expansion compensation errors.

The interferometric values assigned to each block from this analysis form the basis of the NISTmaster gauge block calibration procedure. For each size gauge block commonly in use up to 100mm (4 in.), we have two master blocks, one steel and one chrome carbide. Blocks in the long series,125 mm (5 in.) to 500 mm (20 in.) are all made of steel, so there are two steel masters for each size. Lengths of customer blocks are then determined by mechanical comparison with these masters inspecially designed sequences.

4.4.2 The Comparison Process

There are currently three comparison schemes used for gauge block measurements. The maincalibration scheme uses four gauge blocks, two NIST blocks and two customer blocks, measured sixtimes each in an order designed to eliminate the effects of linear drift in the measurement system. For gauge blocks longer than 100mm (4 in.) there are two different schemes used. For most of theselong blocks a procedure using two NIST blocks and two customer blocks measured four times eachin a drift eliminating design is used. Rectangular blocks over 250 mm are very difficult tomanipulate in the comparators, and a reduced scheme of two NIST masters and one customer blockis used.

Comparators and associated measurement techniques are described in chapter 5. In this chapter, only

49

the measurement assurance aspects of processes are discussed. The MAP for mechanicalcomparisons consists of a measurement scheme, extraction of relevant process control parameters,process controls, and continuing analysis of process performance to determine process uncertainty.

4.4.2.1 Measurement Schemes - Drift Eliminating Designs

Gauge block comparison schemes are designed with a number of characteristics in mind. Unlessthere are extenuating circumstances, each scheme should be drift eliminating, provide one or morecontrol measurements, and include enough repeated comparisons to provide valid statistics. Eachof these characteristics is discussed briefly in this section. The only scheme used at NIST that doesnot meet these specifications is for long rectangular blocks, where the difficulty in handling theblocks makes multiple measurements impractical.

In any length measurement both the gauge being measured and the measuring instrument may changesize during the measurement. Usually this is due to thermal interactions with the room or operator,although with some laser based devices changes in the atmospheric pressure will cause apparentlength changes. The instrumental drift can be obtained from any comparison scheme with moremeasurements than the number of unknown lengths plus one for drift, since the drift rate can beincluded as a parameter in the model fit. For example, the NIST comparisons include sixmeasurements each of four blocks for a total of 24 measurements. Since there are only threeunknown block lengths and the drift, least squares techniques can be used to obtain the best estimateof the lengths and drift. It has been customary at NIST to analyze the measurements in pairs,recording the difference between the first two measurements, the difference between the next twomeasurements, and so on. Since some information is not used, the comparison scheme must meetspecific criteria to avoid increasing the uncertainty of the calibration. These designs are called drifteliminating designs.

Besides eliminating the effects of linear instrumental drift during the measurements, drift eliminatingdesigns allow the linear drift itself to be measured. This measured drift can be used as a processcontrol parameter. For small drift rates an assumption of linear drift will certainly be adequate. Forhigh drift rates and/or long measurement times the assumption of linear drift may not be true. Thelength of time for a measurement could be used as a control parameter, but we have found it easierto restrict the measurements in a calibration to a small number which reduces the interval and makesrecording the time unnecessary.

As an example of drift eliminating designs, suppose we wish to compare two gauge blocks, A andB. At least two measurements for each block are needed to verify repeatability. There are twopossible measurement schemes: [ A B A B ] and [ A B B A ]. Suppose the measurements are madeequally spaced in time and the instrument is drifting by some ∆ during each interval. Themeasurements made are:

50

Measurement Scheme ABAB Scheme ABBA

m1 A A (4.8)m2 B + ∆ B + ∆m3 A + 2∆ B + 2∆m4 B + 3∆ A + 3∆

If we group the measurements into difference measurements:

Scheme ABAB Scheme ABBA

y1 = m1 - m2 = A - (B + ∆) A - (B + ∆) (4.9a)y2 = m3 - m4 = (A+2∆)-(B+3∆) (B + 2∆)-(A+3∆) (4.9b)

Solving for B in terms of A:

B = A - (y1+y2)/2 - ∆ A - (y1-y2)/2 (4.10)

In the first case there is no way to find B in terms of A without knowing ∆, and in the second casethe drift terms cancel out. In reality, the drift terms will cancel only if the four measurements aredone quickly in a uniformly timed sequence so that the drift is most likely to be linear.

A general method for making drift eliminating designs and using the least squares method to analyzethe data is given in appendix A.

The current scheme for most gauge block comparisons uses two NIST blocks, one as a master block(S) and one as a control (C), in the drift eliminating design shown below. Two blocks, one fromeach of two customers, are used as the unknowns X and Y. The 12 measured differences, yi, aregiven below:

y1 = S - Cy2 = Y - Sy3 = X - Yy4 = C - Sy5 = C - Xy6 = Y - C (4.11)y7 = S - Xy8 = C - Yy9 = S - Yy10 = X - Cy11 = X - Sy12 = Y - X

51

(4.11)

The calibration software allows the data to be analyzed using either the S block or the C block as theknown length (restraint) for the comparison. This permits customer blocks of different materials tobe calibrated at the same time. If one customer block is steel and the other carbide, the steel mastercan be used as the restraint for the steel unknown, and the carbide master can be used for the carbideunknown. The length difference (S-C) is used as a control. It is independent of which master blockis the restraint.

There is a complete discussion of our calibration schemes and their analysis in appendix A. Herewe will only examine the solution equations:

S = L (reference length) (4.12a)

C = (1/8) (-2y1 + y2 + 2y4 + y5 - y6 - y7 + y8 - y9 - y10 + y11) + L (4.12b)

X = (1/8) (-y1 + y2 + y3 + y4 - y5 - 2y7 - y9 + y10 + 2y11 - y12) + L (4.12c)

Y = (1/8) (-y1 + 2y2 - y3 + y4 + y6 - y7 - y8 - 2y9 + y11 + y12) + L (4.12d)

∆ = (-1/12)(y1+y2+y3+y4+y5+y6+y7+y8+y9+y10+y11+y12) (4.12e)

The standard deviation is calculated from the original comparison data. The value of eachcomparison, <yi>, is calculated from the best fit parameters and subtracted from the actual data, yi. The square root of the sum of these differences squared divided by the square root of the degreesof freedom is taken as the (short term or within) standard deviation of the calibration:

For example,

∆y12 = (C - B) - y12 (4.13a)

= (1/8) (y2 -2y3 + y5 +y6 + y7 - y8 - y9 - y10 - y11) - y12 (4.13b)

/8)Y( = 2i

12

0i=w

2

52

These deviations are used to calculate the variance.

The standard deviation (square root of the variance), a measure of the short term variation of thecomparison process, is recorded for each calibration in the MAP file, along with other pertinent datafrom the calibration. These data are later used to determine the value of one control parameter, theshort term variation, as described in the next section.

4.4.2.2 Control Parameter for Repeatability

There are a number of ways to monitor the comparator process. The simplest is to compare thevariance of each calibration to an accepted level based on the history. This accepted level (averagevariance from calibration history) is characteristic of the short term repeatability of the comparatorprocess, and is called the "within" variance.

In our system the within variance is used for two purposes. One is as a control point for thecalibration, as discussed below. The second is that being a measure of short term variation of theentire calibration process, it is used to calculate the process uncertainty, a part of the total uncertaintyreported to the customer.

In our system the variance for each calibration is recorded in the MAP file. Each year the last threeyears of data are analyzed to update the process control parameters used for the next year. Figure4.3 shows a typical MAP file data plot.

Figure 4.3. Variation of the short term standard deviation with time for the 0.1 inchsize.

53

The most prominent characteristic of the graph is the continuing reduction of the standard deviationsince 1981. The installation of new comparators during 1987 and the replacement of one of the oldmaster sets with a new chrome carbide set in 1989 have resulted in significant reductions in theprocess uncertainty. When the current process parameter, σw , is calculated, only the data since thelast process change (new master set in 1989) will be used.

If the calibration variance was independent of the block length only one value of the controlparameter would be needed, and the average of the last few years of data from all of the calibrationscould be used as the control point. The variance is, however, made up of many effects, some lengthdependent and some not.

Some of the non-length dependent part of the variance is due to electronic noise in the comparator. Another source is the contact point of the comparator sampling slightly different points on thesurface of the block each time. Due to variations in flatness, parallelism and probe penetration intothe surface, different lengths are measured.

The length dependent part of the variance is due to factors such as thermal variations in the blocks. The variation caused by this factor is not so strong that every block size must be consideredseparately. For example, the effects of thermal variations for 1 mm and 1.1 mm blocks should benearly identical, although the same thermal variation on the 5 mm and 50 mm blocks will bemeasurably different. By separating the blocks into small groups of nearly the same size the numberof measurements is increased, making the determination of the variance more accurate for the groupthan it would be if done on an individual basis.

Each group is made of about 20 sizes, except for blocks over 25 mm (1 in) where each size isanalyzed separately. The groupings in current use are shown in table 4.4.

For each block size the average variance for the last three years is calculated. A weighted averageof the variances from individual blocks is calculated, as shown in eq (4.4).

(4.8)

In the above equation:

σp is the pooled standard deviation of the group,

m is the number of sizes in the group,

2)-n(

S2)-n( =

j

m

1=j

2jj

m

1=jp

2

54

nj is the number of measurements of size j,

sj is the average standard deviation calculated from the MAP file for size j.

The within, σw is calculated for each group.

Table 4.4. Current Groupings for Calculating σw.

Group Number Sizes Units

1 0.010 to 0.09375 inch (thins)2 0.100 to 0.107 inch3 0.108 to 0.126 inch4 0.127 to 0.146 inch5 0.147 to 0.500 inch6* 0.550 to 4.000 inch7* 5.000 to 20.00 inch8 0.0101 to 0.019 inch (.010 thins)9 0.0201 to 0.029 inch (.020 thins)10 0.0501 to 0.059 inch (0.050 thins)11 0.200 to 0.2001 inch (step blocks)12 unassigned13 0.10 to 0.95 millimeter (thins)14 1.00 to 1.09 millimeter15 1.10 to 1.29 millimeter16 1.30 to 1.49 millimeter17 1.50 to 2.09 millimeter18 2.10 to 2.29 millimeter19 2.30 to 2.49 millimeter20 2.50 to 10 millimeter21 10.5 to 20 millimeter22 20.5 to 50 millimeter23* 60 to 100 millimeter24* 125 to 500 millimeter

* NOTE: for sizes over 1 inch and 50 mm each block history is analyzed individually, the groupingshere are merely to simplify the data file structure.

4.4.2.3 Control test for the variance:

To test if the variance of an individual calibration is consistent with the short term variance derivedfrom the calibration history we use the "F-test." The F-Test compares the ratio of these two

55

variances to the F distribution, or equivalently the ratio of the standard deviations squared (variance)to the F distribution. Since the variance is always positive it is a one sided test:

σ2cal/σ2

w < F1-α(k-1,v) (4.16)

where σw is the historical (pooled) short term group standard deviation, σcal is the standard deviationfrom the calibration, α is the upper percentage point of the F distribution with k-1 degrees of freedomin the numerator and v degrees of freedom in the denominator. Using the 12 comparison calibrationscheme, the number of degrees of freedom is 8, i.e. 12 measurements minus the number ofunknowns (S,C,X,Y,∆) plus one for the restraint. The number of degrees of freedom in finding σwfrom the history is generally a few hundred, large enough to use infinity as an approximation. Theα used is 0.01. Using these parameters the control test becomes

σ2cal/σ2

w < 2.5 (4.17)

If this condition is violated the calibration fails, and is repeated. If the calibration fails more thanonce the test blocks are re-inspected and the instrument checked and recalibrated. All calibrations,pass or fail, are entered into the history file. A flag is appended to the record to show that an F testfailed.

4.4.2.4 Control parameter (S-C)

Control parameter σw will monitor certain aspects of the calibration, but not necessarily the accuracy. That is, measurements can have good repeatability and still be wrong. It is possible, for example,that the comparator scale calibration could be completely wrong and the data could still be repeatableenough that the variance (as measured by σw) would be in control. Our second control parametersamples the accuracy of the system by monitoring, for each size, the length difference between twoNIST master blocks (S-C). No matter which master block is taken as the restraint this difference isthe same.

Consider the measurement sequence shown in the previous sections. The customer block lengthsX and Y, as well as the length of one of our master blocks are determined by a model fit to thecomparison data using the known length of the other NIST master block (the restraint). The idea ofthis test is that while the customer block length is not known, the length of our control block isknown and if the calibration process produces the correct length for the control block it probably hasalso produced the correct length for the customer block.

56

figure 4.4. (S-C) data from the MAP files for the 0.1in. blocks from 1982 to 1992. Note the process change in 1988-89 when the "C" block was replaced.

A plot, for the 0.1" size, of all of the S-C measurements from 1982 to the end of 1992 is shown infigure 4.4. There are three distinct control regions shown in the graph by vertical lines. The first,which extends from 1982 to the middle of 1987 shows the performance of the system when we usedtwo steel master blocks and our old (circa 1974) comparators. During the next period, until thebeginning of 1990, we used new comparators with modern electronics. The higher repeatability ofthe new system resulted in a slightly smaller spread in the (S-C) values. In 1989 we replaced the "C"set of steel blocks with chrome carbide, resulting in a shift in the (S-C) values. When we analyzethe MAP data to get the accepted value of S-C to use in the statistical test we obviously ignore alldata before 1989.

Figure 4.5. MAP data for 1.0 in. calibrations from 1982 to 1992. Note that thespread is somewhat larger than that of the 0.1 in size data in figure 4.4 because oftemperature related uncertainties. The C block was replaced in 1989.

57

Figures 4.5 and 4.6 show the corresponding S-C graphs for the 1.0 in. and 10 in. sizes. The mostprominent feature is that the data spread gets larger with length, a reflection of thermal variationsduring the calibrations. The 10 in. blocks show no process jump because the long block comparatorand master blocks have not been changed during the time period shown.

Figure 4.6. MAP data for 10 in. calibrations from 1982 to 1992. There is no processchange for this size because the long block comparator and the two master blockshave not changed during the time period shown.

These charts provide a wealth of information about the calibration. Since the control block is treatedexactly as the customer blocks, the variation in the control block length is identical to long termvariation in the customer block calibrations. The within uncertainty, calculated from deviations fromthe model fit for each calibration only samples the random errors within characteristic times (lessthan 1 or 2 minutes) it takes for a single calibration. The (S-C) history samples both the short termvariations and the long term randomness caused by comparator scale variations, thermometercalibration changes, or operator differences.

The long term value of (S-C) is not necessarily constant for all sizes. Some gauge blocks, asdiscussed earlier, are not stable. If one master block of a size is not stable, the (S-C) recorded foreach calibration will gradually change. This change will be detected when the history for each sizeis analyzed each year to update the control parameters. Usually blocks with large instabilities arereplaced.

Because of possible instability, the long term variation in (S-C) is calculated by the standarddeviation of a linear fit to the (S-C) history data collected from past calibrations. Data for each sizeis fit to the form

58

Li = miT + Ii (4.19)

where Li is the length of the i'th block, mi is the best fit slope, T is the time of the measurement, andIi is the best fit intercept. The standard deviation of the slope, σslope, is determined by the fitprocedure. If the block has a "significant" drift as determined by the test:

mi > 3σslope (4.20)

then the value of the control parameter is taken as the fit value from equation 4.19 when T is equalto half way till the next update. If the slope is not significant the slope is assumed to be zero, andthe data is simply averaged. The standard deviation of the data from the average is then determined.

These deviations are pooled for every block in a group and the pooled standard deviation, σt, isdetermined. This σt is recorded in the calibration master file to be used in the uncertainty calculationreported to the customer.

Finally, as discussed earlier short term variability is characterized by the within standard deviation,σw recorded at each calibration. Long term variability is characterized by σt calculated from the (S-C) history data. These two numbers can be compared and any non-negligible difference can beattributed to uncorrected long term variations or errors in the system. Based on the size of theseuncorrected errors it may or may not be economically useful to search for their sources and reducetheir size.

4.4.2.5 Control test for (S-C), the Check Standard

The length difference between the S and C blocks of each size are called check standards, and theymonitor process accuracy. The control test for accuracy is based on a comparison of the (S-C) valueof the current calibration with the accepted value of (S-C) derived from the (S-C) mechanicalcomparison data that has been accumulated for several years. This comparison is the second controltest for the calibration, and the historically accepted mechanical value of (S-C) is the controlparameter. To be in statistical control, the current calibration must be consistent with the previous populationof measurements. The t-test compares the current measurement of (S-C) to the accepted (S-C) valueusing the standard deviation derived from previous calibrations, σw, as a scale. If the differencebetween the current and accepted value is large compared to this scale the measurement system isjudged to be out-of-control.

The actual test statistic is:

(4.21))(t<Accepted)-(Measured/2

w

59

Where v is the number of degrees of freedom in the accepted total standard deviation, and α is thesignificance level. At NIST we use the 3σ, or 0.3 % level in calculating the total error, and thereforeα=.003 is used in the t-test. The total number of measurements used to calculate the accepted totalstandard deviation is usually over 200, and we can use infinity as an approximation of v.

(4.22)

With these assignments a t-test critical value of 2.62 is used. As a comparison, if v were chosenlower, due to having only a short calibration history, the changes would be small. For v=120 and60 the critical values are only increased to 2.77 and 2.83 respectively.

If this condition is violated the calibration fails, and is repeated. If the calibration fails more thanonce the test blocks are re-inspected and the instrument checked and recalibrated. All calibrations,pass or fail, are entered into the history file. A flag is appended to the record to show that the t-testfailed.

4.4.2.6 Control test for drift

As a secondary result of analyzing comparison data using a drift eliminating design, linear drift ismeasured and can be used as another control parameter. If the drift shows nonlinear behavior intime, the design will not completely eliminate the drift, and the answers will be suspect. However,nonlinear drift will accentuate the deviations from the best line fit of the data and the calibration willfail the F-test. Thus a control based on the drift rate would be redundant. In either case, because ofthe environmental control used in our lab ( 0.1 ºC ) drift has never been significant enough toconsider such a test.

4.4.3 Calculating the total uncertainty

In the general method for combining uncertainties discussed in section 4.3.6 total uncertainty is theroot-sum-of-squares of the uncertainty components times a coverage factor k=2.

For NIST calibrations the total error is made up of random error from the master blockinterferometric history and long term random error of the comparator process derived from controldata. Other sources of uncertainty are the phase correction assigned to each block in theinterferometric measurements and the temperature difference between the customer block and theNIST master block. For blocks over 50 mm the temperature gradients and non-negligible sourcesof uncertainty.. The long term random error of the comparison process is calculated from the (S-C) data collectedin the MAP file. Since each block, master and unknown, is handled and measured in exactly the

2.62<Accepted)-(Measuredw

60

same manner, the variance of the difference (S-C) is a reasonable estimator of the long term varianceof the customer calibration (S-X) or (C-X). The major difference is that since the S and C blocksare of different materials, the variation will be slightly larger than for customer calibrations, forwhich (Standard-X) is always derived from the master block of the same material as the customerblock. Although we expect this difference is small, we have not collected enough customer data toanalyze the difference, and our total uncertainty must be considered slightly inflated.

The total (long term) standard deviation (σtot) is calculated by groups, in much the same manner asσw. One additional step is included because of the possible instability in the length of one of themaster blocks. The (S-C) history of each size is fit to a straight line and the deviations from the lineare pooled with the deviations from the other sizes in the group, as shown in eq (4.15).

In the above equation:

σtot is the pooled long term standard deviation of the group,

m is the number of measurements of all sizes in the group,

Dj is the deviation from the best fit to the (S-C) history data for each size.

As discussed by Cameron [5], the random error for calibration schemes using incomplete blockdesigns is a combination of the short and long term repeatability. In these cases short termrepeatability is the average of the standard deviations of each calibration recorded in the calibrationhistory, σw, and the long term repeatability is σtot. For complete block designs only σtot is needed.

The interferometric uncertainty is essentially a random error. If each block were measured only afew times on a single date the pressure, temperature and humidity calibrations used to makecorrections for the index of refraction and thermal expansion would be sources of systematic error. However, each time our master blocks are re-measured ( 3 year intervals) the pressure, temperatureand humidity sensors are recalibrated. After the blocks have been measured using three or moresensor calibrations, errors of the environmental sensors have been sampled and the variations dueto them have been absorbed into the gauge block calibration data.

The interferometric phase correction has a fairly high uncertainty because of the difficulty of themeasurements. In addition, phase measurement is much too difficult to make on each block and so

(4.23)

2)-(m

D =

2j

m

1=jtot

61

only a few blocks from each manufacturer are tested and the average of these measurements areassigned to all blocks from the same manufacturer. Estimated uncertainty in the phase correctionis around 8 nm ( 0.3 µin).

We have measured temperature gradients on all of our comparators. On short block comparators(<100 mm) we have found gradients up to 0.030 ºC between blocks. Since we do not have extensivemeasurements we consider this as a rectangular distribution of total width 0.030 ºC. Using the rulesof NIST Technical Note 1297 we divide the width by the square root of 3, giving an uncertainty of0.020 ºC. The long block comparators are partially enclosed which reduces the gradients below0.010 ºC. The standard uncertainty from these gradients is only 0.005 ºC.

4.5 Summary of the NIST Measurement Assurance Program

Two major goals of the gauge block MAP program are to provide a well-controlled calibrationprocess to ensure consistently accurate results, and to quantitatively characterize the process so thata rational estimate of process accuracy can be made. Process control is maintained with three testsand data collected from the calibrations is used for both the tests and the estimation of processuncertainty.

In the first test interferometric history of each master block is continually updated and checked forconsistency. Suspicious values are re-measured. As the number of measurements in each historyincrease, the dimensional stability of each block can be determined. If the block is changing length,the projected length at the time of the comparison measurement can be used, or the block can bereplaced. The standard deviation of the history data for each block is used as an estimate of the longterm random uncertainty of the master block value.

In the second test the standard deviation of each mechanical comparison is checked against theaverage value of the standard deviations from previous calibrations. This average standard deviationfor an individual calibration is called the "within" or short term standard deviation (σw). Since theerror sources for similar size blocks are similar, we pool the historical data for groups of similarsized blocks to increase statistical precision. The standard deviation from each calibration is testedagainst the accepted σw by the formula:

σ2cal/σ2

w < 2.5 (4.24)

In the final test each mechanical comparison includes a check standard as well as the master block. The difference between these two NIST blocks, (S-C), is tested against an accepted value obtainedby a linear fit to the recorded (S-C) values for the last few years. The difference between the valueobtained in the calibration and the accepted value is measured against the expected standarddeviation of the measurement, σw, using the formula:

62

(4.25)

The uncertainty reported to the customer is a combination of the type A and type B uncertainties. The major components of uncertainty are the standard deviation of the interferometric history of themaster block used in the calibration, the long term standard deviation of the mechanical comparisonprocess, σtot, and the estimated type B uncertainty. The primary components of type B uncertaintyare the phase correction applied during interferometric calibrations of our master gauge blocks andthe temperature gradients between blocks during the comparison. All of these errors are combinedin quadrature and multiplied by a coverage factor (k) of two, thus arriving at the reported uncertainty.

(4.26)

)(t < Accepted)-(Measured/2

w

B+A k =y Uncertaint Total 22

63

5. The NIST Mechanical Comparison Procedure

5.1 Introduction

Techniques described here for intercomparing gauge blocks with electro-mechanical gauge blockcomparators are those used at NIST and are the results of considerable experimental work, statisticaldesign and process analysis. Some laboratories will find these techniques ideally suited to theirneeds while others may wish to modify them, but in either event the process control methodsdescribed here will promote the analysis of measurement requirements and the application ofstatistical principles.

Length values are assigned to unknown gauge blocks by a transfer process from blocks of knownlength. Briefly, the transfer process is a systematized intercomparison of 3 or 4 blocks of the samenominal size using an electro-mechanical comparator. One or two blocks are unknowns and two arestandards. Such a process is ideally suited to comparing objects with nearly identical characteristics,such as gauge blocks, and the process provides the redundancy needed for statistical analysis.

We will also describe the corrections needed for the case where the blocks are not made of identicalmaterials. The correction methods apply to all sizes but special precautions will be described forlong gauge blocks where temperature effects predominate. When the master and unknown blocksare of the same material, which for our calibrations include both steel and chrome carbide blocks,no corrections are needed.

5.2 Preparation and Inspection

It is essential that the gauge blocks being calibrated, as well as the standards being used, be clean andfree from burrs, nicks and corrosion if the calibration is to be valid. Therefore, it is necessary toinstitute an inspection, cleaning and deburring procedure as part of every calibration. Such aprocedure will also protect the comparator working surface and the diamond styli.

5.2.1 Cleaning Procedures

Gauge blocks should be cleaned in an organic solvent such as mineral spirits. Solvents such astrichloroethylene or benzene should be avoided because of health risks involved with their use. Freons should not be used because of environmental hazards they pose. The bottom of the bathshould be lined with a waffled mat of neoprene rubber to minimize scratches and burrs to the gaugeblock surfaces caused by grit or metal to metal contact.

Rectangular blocks should be immersed in the bath and a soft bristle brush used to remove the layerof protecting oil or grease from all surfaces. Lint-free towels should be used to dry the blocks. Asecond cleaning using ethyl alcohol is recommended to remove any remaining residue. Again, wipewith lint-free towels.

Square type blocks are cleaned in the same manner as the rectangular blocks except that the center

64

hole needs special attention. The hole must be thoroughly cleaned and dried to prevent "bleedingout" of an oil film onto the contact surfaces and supporting platen when the block is positionedvertically. A .22 caliber gun cleaning rod makes an ideal instrument for cleaning the bore of blocksin the size range of 2 mm to 500 mm. A solvent moistened piece of lint-free paper towel can bepushed through the hole repeatedly until all visible residue is removed.

5.2.2 Cleaning Interval

It is recommended that the solvent and alcohol cleaning process be completed several hours priorto use because the cleaning process will disturb the gauge block thermal state. If blocks are to bejoined together or to platens by wringing, recleaning is necessary prior to introducing a wringing filmto ensure that no foreign substances are on the wringing surfaces that could corrode or mar thecontact surfaces.

5.2.3 Storage

Cleaned blocks should be stored in trays covered with lint-free paper or on temperature equalizationplates that are clean. A cover of lint-free towels should be used to secure the blocks from dust andother airborne solids and liquids. When blocks are not in daily use, they should be coated with anyone of the many block preservatives recommended by the gauge block manufacturers. It isrecommended that steel gauge blocks be coated to prevent rust when the work area relative humidityexceeds 50%.

Foodstuffs should be banned from all calibration areas, as acids and salts will corrode steel gaugesurfaces. Hands should be washed to reduce corrosion on blocks when gloves or tongs cannot beused conveniently.

5.2.4 Deburring Gauge Blocks

Gauge blocks can often be restored to serviceable condition by stoning if the damage is not severe.Three different optically flat gauge block stones are available: black granite, natural Arkansas, orsintered aluminum oxide. All stones are used in the same manner, but are used for different purposesor on different materials.

The black granite deburring stone is used on steel blocks to push a burr of metal back into the areafrom which it was gouged. The natural Arkansas deburring stone is a white fine-grained stone thatis used on steel blocks to remove burrs and nicks caused by everyday mishaps. It is especiallyeffective on damaged edges of gauging faces. The sintered aluminum oxide stone is used to treateither steel or carbide gauge blocks. It is most effective on hard materials such as tungsten andchromium carbide.

The stone and block should both be cleaned with alcohol before stoning. Sweep the surfaces witha camel's hair brush; then slide the damaged surface of the block onto the stone, taking precautionsthat no foreign matter is trapped between the two surfaces, and with firm downward pressure the

65

block is moved over the stone in either a circular or a back and forth motion until the metal isrestored to its original placement. Generally, there is a lessening of drag between the two surfacesas the burr is reduced.

After stoning and cleaning a block, one can check the surface by wringing the restored surface to afused silica flat. Looking through the back side of the flat, the interface between the block and flatshould appear uniformly grey when it reflects white light. Any light areas or color in the interfaceindicates poor wringing caused by a burr that has not been fully reduced to the normal gauge surface.

Periodically, depending on use, the stone should be cleaned with a stiff natural fiber brush, soap andwater to remove embedded pieces of metal and other foreign matter that contaminates the surfacewith use. Allow moisture absorbed by the stone surface to evaporate before using the stone.

5.3 The Comparative Principle

A comparator is a device for measuring differences. The unknown length of a block is determinedby measuring the difference between it and a reference block of the same nominal size and thencalculating the unknown length. Let Lc equal the comparator length when its indicator reads zero(the zero may be at the left end or the center of the scale). Then, by inserting the two blocks oflength Lx (unknown) and Lr (reference) and reading numerical values x and r from the comparatorscale, we get

Lx = Lc + x (5.1)Lr = Lc + r (5.2)

Solving these equations for Lx:

Lx = Lr + (x-r). (5.3)

Note that the quantity (x-r) is the difference between the two blocks. If the blocks are made ofdifferent materials we add a quantity, δ, to each comparator reading to compensate for elasticdeformation of the contact surfaces. We must also add a correction for the thermal expansion if thecalibration temperature T is not 20 ºC. The equation becomes

Lx = Lr + (x - r) + (δx - δr) + L(αr - αx)(t-20). (5.4)

The quantity (δx - δr) in this equation is the deformation correction, Cp. αx and αr are the thermalexpansion coefficients, and L the blocks' nominal lengths. Using Ct for this thermal correction theequation becomes

Lx = Lr + (x-r) + Cp + Ct. (5.5)

66

If two probes are used, one lower and one upper, there are two penetration corrections and Cp mustcontain a lower and upper component.

A key result of the comparative principle is that if the two blocks being compared are the samematerial, Cp and Ct are zero.

5.3.1 Examples

1. A single probe comparator is used to measure a 50 mm tungsten carbide gauge block witha 50 mm steel reference standard gauge block. The measurement data are:

Stylus force (top): 0.75 NStylus force (bottom): 0.25 NStylus tip material: diamondStylus tip radius: 6 mmLr of reference at 20 ºC 50.00060 mmTemperature of blocks: 20.4 ºC Expansion coefficient:

Steel 11.5x10-6/ ºC Tungsten carbide 6 x10-6/ ºC

Comparator reading on tungsten block 1.25 µmComparator reading on steel block 1.06 µm

The unknown length (Lx) at 20 ºC can be found using the last equation and the penetrationcorrections from table 3.4:

Penetration, µm (µin)Steel 0.14 (5.7)Tungsten Carbide 0.08 (3.2)

Then

Lx = 50.00060 + (1.25-1.06)x10-3 + 50x(11.5-6)x10-6x(20.4 - 20) + (0.08 - 0.14)x10-3 (5.6a)

Lx = 50.00060 + 0.00019 + 0.00011 - 0.00006 (5.6b)

Lx = 50.00084 mm (5.6c)

67

2. A dual probe comparator is used to measure a 10 mm steel block with a 10 mm chromecarbide reference block. From the following data find the length of the steel block at 20 ºC.

Stylus force: 0.75 N upper, 0.25 N lowerStylus tip material: diamondStylus tip radius: 6 mmLr of reference at 20 ºC 9.99996 mmTemperature of blocks: 23.6 ºC Expansion coefficient:

Steel 11.5x10-6/ ºC Chrome carbide 8.6x10-6/ ºC

Comparator reading on chrome block 0.00093 mmComparator reading on steel block 0.00110 mm

penetration (0.25 N) penetration (0.75 N) TOTALSteel .07 µm 0.14 µm 0.21 µmChrome Carbide .06 µm 0.12 µm 0.18 µm

Solution:

measurements thermal correction penetration correction

Lx = 9.99996+(.00110-.00093) +10x(8.6-11.5)x10-6x(23.6 - 20) + (.21-.18)x10-3 (5.7a)

Lx = 9.99996 + .00017 - .00010 + 0.00003 (5.7b)

Lx = 10.00006 mm (5.7c)

In summary, deformation corrections can be easily and accurately applied if the operative parametersof probe tip radius and probe force are known and if the probe tip is spherical and remainsundamaged. Extremely low forces should be avoided, however, because poor contact is a risk. Itis also important to know that thermal conditions, if ignored, can cause far greater errors thandeformation, especially for long lengths.

Currently the NIST comparator program uses steel master blocks for steel unknowns and chromecarbide masters for chrome or tungsten carbide unknowns. This allows the penetration correctionto be ignored for all of our calibrations except for a very few sets of tungsten carbide and othermaterials that we receive. Because of the small number of gauge blocks sent for calibration it is noteconomically feasible to have master sets for every possible gauge block material. For these blocks

68

the correction is minimized by the use of the master blocks with the most closely matched elasticproperties.

5.4 Gauge Block Comparators

There are a number of suitable comparator designs and a typical one is shown schematically in figure5.1. An upper stylus and a lower stylus contact the gauging faces of a block supported on an anvil. Each stylus is attached to a linear variable differential transformer (LVDT) core. The amplifierdisplays the difference between the two LVDT signals.

Figure 5.1. Schematic drawing of a gauge block comparator showing the component parts.

The NIST comparators have a more complex mechanical arrangement, shown in figure 5.2. The useof pivots allows measurement of the relative displacement of two measuring styli using only oneLVDT. Since one of the main causes of variations in measurements is the LVDT repeatability,reducing the number of LVDTs in the system enhances repeatability.

69

Figure 5.2. NIST gauge block comparator.

An important comparator feature is that the block is held up by the anvil (also called platen),removing the weight of the block from the lower stylus. The force of the gauging stylus on thebottom of the block is therefore the same for every block measured. Were this not the case thecorrections needed for deformation at the point of contact would depend on the weight of the blockas well as it's material. There are, however, systems that support the block on the lower stylus. Inthese cases a penetration correction which depends on the material, size and geometry of the blockmust be made for each block.

Since the LVDT range is small, the height of one of the sensors must be mechanically changed foreach size gauge block. The highest feature on figure 5.2 is the crank used to rotate a fine pitch screwattached to the top stylus. For each size the top stylus is set and locked so that the LVDT output ison the proper scale. The lifting lever raises the top stylus from the block and locks the pivot systemso that the stylus cannot crash into the gauging surface by mistake during the gross adjustment to thenext size.

For the most precise measurements, it is important that the defined gauge point of a block becontacted by the comparator stylus. If the two gauging surfaces are not exactly parallel the measuredblock length will vary with the gauging point position. This is particularly important for long blocks,which tend to have less parallel gauging surfaces than short blocks. To ensure that contact will bemade at the proper point a metal or plastic bar about 1/4 inch thick is fastened to the anvil behindthe stylus and is positioned to stop the gauge blocks so the stylus will contact the gauge point. Thestop used on our long block comparator is shown in figure 5.3.

70

Figure 5.3. The mechanical stop is used to force the gauge block toa nearly identical position above the stylus for each measurement.

The LVDT signal (or difference between two LVDTs) is displayed on a meter graduated in lengthunits. The analog output is also read by a digital voltmeter which, when triggered by a foot switch,sends the measured voltage to a microcomputer. The difference in length between two gauge blocksis obtained by inserting the blocks, one at a time, between the stylus tips and taking the differencebetween the two readings.

5.4.1 Scale and Contact Force Control

The comparator transducer must be calibrated to relate output voltage to probe displacement fromsome arbitrary zero. Our comparators are calibrated regularly using a set of 11 gauge blocks thatvary in length from 0.200000 inches to 0.200100 inches in 10 microinch increments. Eachcalibration is recorded in a probe calibration file and the last 20 calibrations are available to theoperator as a control chart with appropriate control limits. If a probe calibration is inconsistent withthe previous calibrations the instrument is examined to find the cause of the discrepancy.

The NIST comparators for blocks less than 100 mm long use contact forces of 0.75 N (3 ounces) forthe upper contact and 0.25 N (1 ounce) for the lower contact. The corresponding forces for the longgauge block comparator (over 100 mm) are 1.2 N (5 oz) and 0.75 N (3 oz). The force on both upperand lower contacts are checked daily using a calibrated, hand held, force gauge. If the force is foundto be in error by more than 0.1 N, adjustments are made.

5.4.2 Stylus Force and Penetration Corrections

The penetration of the spherical stylus into the planar gauge block has already been presented in table3.4. Values are correct to a few nanometers. Calculation of the value assumes a perfectly sphericalsurface with a nominal radius of curvature of 3 mm. This assumption is not always true. Effects of

71

deviations from the assumed probe geometry can be significant and have plagued comparisonprocedures for many years.

Usually the probe is not exactly spherical, even when new, because of the characteristics of diamond. The probe is manufactured by lapping, which if diamond were an isotropic media would producea spherical form. Diamond, however, is harder in some directions than in others. The lappingprocess removes more material in the soft directions producing an ellipsoidal surface. Even worse,the diamond surface slowly wears and becomes flattened at the contact point. Thus, even though thecalculation for deformation of a spherical-planar contact is nearly exact, the real correction differsbecause the actual contact is not a perfect sphere against a plane.

In the past these effects were a problem because all NIST master blocks were steel. As long as thecustomer blocks were also steel, no correction was needed. If, however, other materials weremeasured a penetration correction was needed and the accuracy of this correction was a source ofsystematic error (bias).

A gauge block set bias is determined from the history of each block. First, the yearly deviation ofeach block is calculated from its historical average length. Then the deviations of all the blocks arecombined by years to arrive at the average change in the length of the set by year. This averageshould be close to zero, in fact if there are 81 blocks in the set the average will, if in control, fallwithin the limits 3σ/ 81 or σ/3. The method of examining set averages is useful for many types ofsystematic errors and is more fully described in Tan and Miller [36].

Figure 5.4 shows the set bias for three customer gauge block sets having lengthy histories. Two sets, represented by squares are steel and the third, represented by asterisks, is chrome carbide.

Figure 5.4. Bias of three gauge block sets since 1960. The squares are steel blocksand the crosses are chrome carbide blocks.

72

The two steel set calibration histories are considerably better than those for the chrome carbide. Also, before the early 1970's all calibrations were done by interferometry. The chrome carbidecalibrations have degenerated since 1970 when mechanical comparisons were introduced as thecalibration method.

In the past it was believed that carbide calibrations were not as accurate as steel and one extramicroinch of uncertainty was added to carbide calibrations. The graph shows that this belief gainedfrom experience was accurate. The NIST calibration procedures were changed in 1989 to eliminatethis systematic error by having one set of master blocks made from chrome carbide. When steelblocks are measured the steel master is used as the reference block and when carbide blocks aremeasured the chrome carbide block is used as the reference. In each case the difference in lengthbetween the steel and carbide masters is used as the check standard.

5.4.3 Environmental Factors

A temperature controlled laboratory is essential for high precision intercomparisons. The degree oftemperature control needed depends on the length of the blocks being compared, differences inthermal expansion coefficients among the blocks, and the limiting uncertainty required for theunknown blocks. At NIST, the gauge block calibration areas are specially designed laboratories withtemperature control at 20.1 ºC. Special precautions, described in this chapter, are taken for blocksover 100 mm.

5.4.3.1 Temperature Effects

The correction equation for thermal expansion is

δL/L = α(20-T) (5.8)

Where L is the lenght of the block, δL the change in length due to thermal expansion, α is thethermal expansion coefficient, and (20-T) the difference between the measurement temperature and20 ºC.

When two blocks are compared, the length difference (m=S-X) between the blocks may needcorrection due to the difference in thermal expansion coefficients and temperatures of the blocks. If corrected the calibration result becomes

Xcal = Sref - m + δL(thermal) (5.9)

δL/L = (αX-αS)(20-TS)+αX(TS-TX) (5.10)

73

Uncertainty of the calculated δL arises from uncertainty in α, ∆T, and the generally unmeasuredtemperature difference between S and X. An estimate of the upper limit of this uncertainty is:

Uncertainty = L(αX-αS)(δTX) + L(δαS+δαX)(20-TS) + LαX(TS-TX) (5.11)

where δα is the uncertainty in the thermal expansion coefficient and δT is the uncertainty in thetemperature measurement.

We illustrate the effects with a simple example. Three 250 mm gauge blocks are beingintercompared, a steel block is the reference master designated S, and one steel block and onechrome carbide block are designated X and Y. The thermal expansion coefficients and relateduncertainties are:

S: (11.5 0.1) x 10-6/ºC X: (11.5 0.5) x 10-6/ºC Y: ( 8.4 0.5) x 10-6/ºC

The blocks are measured at 20.5 ºC with an uncertainty of 0.2 ºC, and from previous practice it isknown that there is the possibility of a temperature difference between blocks of up to 0.1 ºC.

Length errors in three categories can be illustrated with these data: those caused by (1) temperaturemeasurement uncertainty, α δT; (2) gauge block expansion coefficient uncertainty, Lδα(t-20); and(3) temperature differences between the reference and unknown blocks.

(1) The temperature measurement uncertainty of 0.2 ºC together with the expansion coefficientsof the unknown blocks yields:

Uncertainty = L(αX-αS)δT (5.12)

(2) The uncertainty in the thermal expansion coefficients together with a measurementtemperature different than 20 ºC yields:

Uncertainty = L(δαS+δαX) |20-TS | (5.15)

U( X or Y) = 250 x (0.1 x 10-6 + 0.5 x 10-6) x 0.5 = 0.00007 mm (3 µin) (5.16)

(3) If the master and unknown blocks are not at the same temperature, there will be an errordepending on the size of the temperature gradient:

74

Uncertainty = L αX (TS-TX) (5.17)

U(X) =250 x 11.5 x 10-6 x 0.1 = 0.0003 mm (12 µin) (5.18)

U(Y) = 250 x 8.4 x 10-6 x 0.1 = 0.0002 mm (8 µin) (5.19)

The example shows that the uncertainty in the thermal expansion coefficient is not important if thetemperature is kept close to 20 ºC. The uncertainty in the temperature measurement is veryimportant for blocks longer than 25 mm, and is critical for blocks longer than 100 mm. Finally,temperature differences between blocks are a serious matter and measures to prevent these gradientsmust be taken. Ways of controlling temperature effects are discussed in the next section.

5.4.3.2 Control of Temperature Effects

In the NIST comparison process it is assumed that the gauge blocks are all at the same temperature,i.e. the temperature of the stabilization plate or comparator anvil. A large error in the comparisonprocess can be introduced by temperature gradients between blocks, or even within a single block. For example, a temperature difference of 0.5 ºC between two 100 mm steel blocks will cause anerror of nearly 0.6 µm (24 µin) in the comparison. Two causes of temperature differences betweenblocks are noteworthy:

1. Room temperature gradients from nearby heat sources such as electronic equipment cancause significant temperature differences between blocks even when they are stored relativelyclose to each other before comparison.

2. Blocks with different surface finishes on their non-gauging faces can absorb radiant heatat different rates and reach different equilibrium temperatures.

The magnitude of these effects is proportional to gauge block length.

A number of remedies have been developed. For short blocks (up to 50 mm) the remedies are quitesimple. For example, storing the blocks, both standards and unknowns, on a thermal equalizationplate of smooth surface and good heat conductivity close to the comparator but away from heatsources. Also, the use of insulated tweezers or tongs to handle the blocks and a systematic, rhythmicblock handling technique in the comparison procedure to ensure identical thermal environment foreach block. The use of a special measurement sequence called a drift eliminating design is discussedin section 5.6.

Precautions for long blocks (over 50 mm) are more elaborate. At NIST the comparator is enclosedin an insulated housing partially open on the front to allow handling the blocks with tongs. Blocksare stored on the comparator anvil and each block is wrapped in three layers of aluminized mylar toassure that each block has the same reflectivity to incident radiation. The insulation, although very

75

thin, also limits the largest heat transfer mechanism, convection, during the measurement. Thisreduces thermal changes due to operator presence during calibration.

Laboratory lights remain off except for one lamp well removed from the comparator, but givingenough illumination to work by. The visible light is a secondary radiation problem, the infraredradiation from the lights is the primary problem. Heat sources are kept as far away from thecomparator as possible and the comparator is located away from other activities where laboratorypersonnel might congregate. As a further precaution during intercomparisons the operator wears acape of Mylar reflective film and heavy cotton gloves while handling the blocks with tongs. Finally,as with short blocks, the handling procedure is systematic, rhythmic and quick but not rushed. Rectangular blocks longer than 100 mm are measured using a reduced number of intercomparisonsbecause of difficulty in handling this type of long block with its tendency to tip on its narrow base.

Temperature problems can be detected in long block intercomparisons by reversing the storagepositions of the 4 blocks and repeating the intercomparison after suitable equalization time. Temperature gradients in the storage area will be revealed by a significant change in the relativelengths of the blocks. Still another method is to measure the temperature of each block with athermocouple. A simple two junction thermocouple used with a nanovoltmeter can measuretemperature differences of 0.001 ºC or less. Tests should be made for gradients between blocks andinternal gradients (top to bottom) within each block. When a gradient free environment isestablished a thermometer mounted on the comparator will be adequate for detecting unusual orgross changes.

Relative humidity is held below 50% to prevent corrosion of blocks and instruments. A fulldiscussion of environmental requirements for calibration laboratories can be found in the ANSIstandard B89.6.2.

5.5 Intercomparison Procedures

Once the blocks have been cleaned, as described in section 5.2 they are laid out on an aluminumsoaking plate to allow all of the blocks to come to the same temperature. The blocks are laid out in4 parallel rows in each size sequence, each row being a different set. The two NIST master blocks,S and C, are the top and second row respectively because each comparison scheme begins S-C. Thecustomer sets, X and Y are set out as the third and fourth rows. It is not necessary to use this specificsetup, but as in all facets of calibration work the setup should be exactly the same for eachcalibration to reduce errors and allow the operators to evolve a consistent rhythm for the comparisonprocedure.

The length of time needed for blocks to thermally equilibrate varies depending on block size and thecontact area with the soaking plate. For blocks less than 10 mm long the same time can be used forall sizes. For longer sizes the contact area is important. If long blocks are laid down so their entireside is resting on the soaking plate then they will equilibrate much like the short blocks. If the blocksare wrapped in mylar this is no longer the case because the mylar and rubber bands used to hold the

76

mylar on insulate the blocks from the plate. Wrapped blocks thus must stand upright on the plate. Since the thermal mass of the block increases but the contact area does not, the longer the block, thelonger time it will take to come to thermal equilibrium. The minimum times used in our lab areshown in table 9.

Table 5.1

Minimum Thermal Soaking Times in Minutes

Length (metric) Length(English) Time (min)

0 - 10 mm 0 - 0.25" 30 10 - 25 mm 0.3 - 1.0" 60 25 - 100 mm 2.0 - 4.0" 100125 - 500 mm 5.0 - 20" 8 h

Experiments establishing optimum equalization times should be conducted in your own laboratorybecause of the many variables involved and differing measurement uncertainty requirements.

5.5.1 Handling Technique

The success of any intercomparison scheme is largely dependent on block handling techniques. These techniques include the insertion of all blocks between the styli in a like manner. The operatorshould develop a rhythm that, after practice, will ensure that each block is handled for approximatelythe same length of time. The time to make 24 observations should be approximately 2 to 4 minutesfor an experienced operator with automatic data recording or an assistant doing the recording.

A camel hair brush or an air bulb is useful for sweeping or blowing dust particles from the blocksand the anvil just before insertion.

Short blocks are moved about by grasping them with rubber tipped tweezers or tongs. For handlingsquare style blocks, the tips of the tweezers may be bent to accommodate this configuration.

For added insulation, light cotton gloves may be worn in addition to using tweezers when handlingblocks from 10 mm to 100 mm (0.4 inch through 4 inches). Blocks above the 100 mm size arecompared on a different instrument situated in a more precisely controlled temperature environmentand the operator may wear a Mylar reflective cape as well as heavy gloves. A special pair of tongsis used to facilitate moving long blocks about the anvil and between the measuring probes. Thedesign used is shown in figure 5.5. Operators have found that the least awkward place to grasp longblocks is about 10 mm above the bottom.

For our long block comparator the anvil stop (see figure 5.3) plays an important part in seating theprobe on the block surface as well as positioning the gauge point between the probes. The block is

77

Figure 5.5. Tongs for moving long square cross section gauge blocks.

moved tangentially toward the stop allowing one corner to touch the stop and then with an angularmotion (which seats the probe on the block surface) proceed to abut the entire edge of the block tothe stop.

5.6 Comparison Designs

5.6.1 Drift Eliminating Designs

The generation and analysis of drift eliminating designs is described in appendix A. In this sectionwe will discuss the four designs used for gauge block calibrations at NIST. The normal calibrationscheme is the 12/4 design that uses two master blocks and two customer blocks. The design consistsof all 12 possible comparisons between the four blocks. When only one set of customer blocks isavailable we use the 6/3 design which uses two master blocks and one customer block. The designconsists of all six possible comparisons between the three blocks.

There are also two special designs used only for long gauge blocks. For long square cross sectionblocks (Hoke blocks) we use the 8/4 design which uses 8 comparisons among four blocks; twomaster blocks and two customer blocks. Long rectangular blocks are difficult to manipulate,particularly when over 200 mm. For these blocks we use a simplified measurement scheme usinga master, a check standard, and one customer block in 4 measurements. For historical reasons thisscheme is called "ABBA."

5.6.1.1 The 12/4 Design

Gauge blocks were calibrated using the 8/4 design for nearly 15 years. It was replaced in 1989 witha new design, the 12/4 design. The new scheme using 12 intercomparisons replaced the earlierscheme which used 8 intercomparisons. Increasing from 8 to 12 intercomparisons evolved from adetermination to eliminate the systematic error caused by the varying contact deformation notedearlier. This design was described in detail in section 4.4.2.1. Here we will only discuss the reasonfor its development.

78

Previously, the restraint was the sum of the known lengths of block S and C, (S+C). The benefit ofthis restraint was that the percentage uncertainty is proportional to the uncertainties of the two blocksadded together in quadrature. Since the uncertainty in S and C were generally about the same, theuncertainty in the restraint was reduced by a factor of 2. The drawback was that if the S, C, andcustomer blocks were not all the same material, corrections for the difference in stylus penetrationwas required. As discussed in the previous section this problem was found to result in measurablesystematic errors. A study of our calibration history showed that contact pressure variations and thechanging stylus geometry due to wear could not be controlled well enough to prevent nearly 25 nmof additional calibration uncertainty.

To reduce this problem one set of steel masters was replaced with chrome carbide masters. A newscheme was then designed using only the steel masters as the restraint for steel customer blocks, andthe carbide masters as the restraint for carbide customer blocks. This eliminates the penetrationcorrection when steel or chrome carbide blocks are calibrated, but it also increases the uncertaintyin the restraint. To lower the total uncertainty the number of measurements was increased, offsettingthe increased uncertainty in the restraint by reducing the comparison process random error .

5.6.1.2 The 6/3 Design

On some occasions two sets with the same sizes are not available for calibration. In the past the oneset was used as both X and Y, giving two answers for each block, the average of the two being usedfor the report. To eliminate excess measurements a new design for one customer block wasdeveloped. The six comparisons are given below:

y1 = S - Cy2 = X - Sy3 = C - X (5.20)y4 = C - Sy5 = X - Cy6 = S - X

When the length L is the master block, the lengths of the other blocks are given by the followingequations. The parameter ∆ is the length drift/(time of one comparison).

S = L (5.21a)

C = (1/6)(-2y1 + y2 + y3 + 2y4 - y5 - y6) + L (5.21b)

X = (1/6)(-y1 + 2y2 - y3 + y4 + y5 - 2y6) + L (5.21c)

<∆> = (1/6)(-y1 - y2 - y3 - y4 - y5 - y6) (5.21d)

79

Formulas for the standard deviation for these quantities are in appendix A.

5.6.1.3 The 8/4 Design

The original 8/4 comparison design is currently only used for long rectangular blocks (100 mm andlonger). This original design was drift eliminating but one comparison was used twice and therewere several consecutive measurements of the same block. In the current (and new) 8/4 scheme, likethe new 12/4 design, the length of only one block is used as the restraint, resulting in a slightly largeruncertainty. For long blocks the 12/4 design is not appropriate because the added time to make theextra comparisons allows drift to become less linear, an effect which increases the apparent randomcomponent of uncertainty.

The current 8/4 design is as follows:

y1 = S – C y2 = X - Yy3 = Y – S y4 = C - X (5.22)y5 = C – S y6 = Y - Xy7 = S – Y y8 = X – C

For the case where S is the master block (known length) the values of the other blocks are:

S = L (5.23a)

C = (1/12) (-4y1 - 2y3 + 2y4 + 4y5 - 2y7 + 2y8) + L (5.23b)

X = (1/12) (-2y1 + 2y2 + 3y3 + 3y4 + 2y5 - 5y7 - y8) + L (5.23c)

Y = (1/12) (-2y1 - 2y2 + y3 + 5y4 + 2y6 - 3y7 - 3y8) + L (5.23d)

∆ = (-1/8)(y1 + y2 + y3 + y4 + y5 + y6 + y7 + y8) (5.23e)

5.6.1.4 The ABBA Design

The actual measurement scheme for this design is X-S-C-X, where X is the customer block, S andC are the NIST standard and control blocks. The name ABBA is somewhat of a misnomer, but hasbeen attached to this design for long enough to have historical precedence.

The design is used only for long (over 200 mm or 8 in) rectangular blocks. These blocks have a verysevere aspect ratio (height/width) and are very difficult to slide across the platen without risk oftipping. The X block is set on the anvil and under the measurement stylus when the blocks are setup the evening before they are measured. The S and C master blocks are also arranged on the anvil. The next morning the X block need only be move twice during the measurement, reducing thepossibility of tipping. The measurement is repeated later and the two answers averaged. Obviously

80

the F-test cannot be used, but the value of S-C is used as the control in a manner similar to the othermeasurement schemes.

5.6.2 Example of Calibration Output Using the 12/4 Design 1 GAUGE PROGRAM Tuesday September 10, 1991 11:01:2823 Calibration: 4.000000 Inch Gauge Blocks Observer: TS4 Block Id's: S: 4361 C: 1UNG X: XXXXXXX Y: YYYYY Federal: F4 56 Observations Observations 7 Blocks at Ambient Conditions at Standard Conditions Differences Residuals8 ----------------------------------------------------------------------------------------------------------------9 S ** C 52.88 62.51 66.28 72.92 -6.64 -1.0010 Y ** S 62.50 53.32 72.91 66.71 6.20 .2211 X ** Y 63.82 62.28 77.21 72.68 4.52 1.0712 C ** S 62.18 54.15 72.59 67.54 5.05 -.5813 C ** X 62.19 63.36 72.60 76.75 -4.15 -.3514 Y ** C 63.15 62.26 73.55 72.66 .89 .5515 S ** X 53.87 62.69 67.26 76.09 -8.80 .6216 C ** Y 62.06 62.19 72.46 72.60 -.14 .2217 S ** Y 53.57 62.87 66.96 73.27 -6.32 -.3318 X ** C 62.76 62.21 76.15 72.62 3.53 -.2719 X ** S 62.86 53.77 76.25 67.16 9.09 -.3420 Y ** X 62.49 62.77 72.90 76.17 -3.27 .192122 Obs Within S.D.= .68 Obs Control= -5.64 F-test = 6.8923 Acc Within S.D.= .26 Acc Control= -8.90 T-test = 9.0724 Acc Group S.D.= .36 Temperature= 19.712526 Deviation27 Serial Nominal Size from Nominal Total Uncertainty Coef. Block28 Number (inches) (microinch) (microinch) ppm/C Material29 ------ ------------ ------------ ----------------- ----- -------------30 4361 4.000000 -8.32 1.86 11.50 steel31 1UN6 4.000000 -2.68 1.86 8.40 chrome carbide32 MARTMA 4.000000 -1.86 1.89 11.50 steel33 TELEDY 4.000000 0.65 2.53 8.40 chrome carbide3435 Process not in statistical control ...

Comments referenced by line numbers:------------------------------------------------------------------------------1. The header shows which computer program was used (GAGE), the date and time of thecalibration. The date is recorded in the MAP file.

3. Block size and operator are identified and later recorded in the MAP file.

4. Block IDs are given here. The S and C blocks are NIST master blocks. The X and Y

81

blocks are shortened forms of customer company names, represented here by strings of Xsand Ys. The ID of the comparator used in the calibration, F4, is also given and recorded inthe MAP file.

9-20. These lines provide the comparison data at ambient temperature and corrected to 20ºC using the thermal expansion coefficients in lines 30-34, and the temperature recorded inline 24. The differences are those between the corrected data. In the final column are theresiduals, i.e., the difference between the best fit to the calibration data and the actualmeasured values.

22-25 These lines present statistical control data and test results.

The first column of numbers shows observed and accepted short term (within) standard deviation. The ratio of (observed/accepted) squared is the F-test value, and is shown in column 3. The lastnumber, group S.D. is the long term standard deviation derived from (S-C) data in the MAP file.

The top line in the second column shows the observed difference (S-C), between the two NISTmasters. The second line shows the accepted value derived from our history data in the MAP file. The difference between these two numbers is compared to the long term accepted standard deviation(group S.D. in column 1) by means of the t-test. The ratio of the difference and the group S.D. is thevalue of the t-test shown in column 3.

30-33 These lines present the calibration result, gauge block material and thermal expansioncoefficient used to correct the raw data. Note that in this test the two customer blocks are of differentmaterials. The same data is used to obtain the results, but the length of the steel customer block isderived from the NIST steel master and the length of the chrome carbide customer block is derivedfrom the NIST chrome carbide master.

35 Since the F-test and t-test are far beyond the SPC limits of 2.62 and 2.5 respectively, thecalibration is failed. The results are not recorded in the customer file, but the calibration is recordedin the MAP file.

The software system allows two options following a failed test: repeating the test or passing on tothe next size. For short blocks the normal procedure is to reclean the blocks and repeat the testbecause a failure is usually due to a dirty block or operator error. For blocks over 25 mm a failureis usually related to temperature problems. In this case the blocks are usually placed back on thethermal soaking plate and the next size is measured. The test is repeated after a suitable time for theblocks to become thermally stable.

5.7 Current System Performance

For each calibration the data needed for our process control is sent to the MAP file. The informationrecorded is:

82

1. Block size2. Date3. Operator4. Comparator5. Flag for passing or failing the F-test, t-test, or both6. Value of (S-C) from the calibration7. Value of σw from the calibration

Process parameters for short term random error are derived from this information, as discussed inchapter 4 (see sections 4.4.2.1 and 4.4.2.2). The (S-C) data are fit to a straight line and deviationsfrom this fit are used to find the standard deviation σtot (see 4.4.3). This is taken as an estimate of

long term process variability. Recorded values of σw are averaged and taken as the estimate ofshort term process variability. Except for long blocks, these estimates are then pooled into groupsof about 20 similar sizes to give test parameters for the F-test and t-test, and to calculate theuncertainty reported to the customer.

Current values for these process parameters are shown in figure 5.6 (σtot) and table 5.2. In generalthe short term standard deviation has a weak dependence on block length, but long blocks show amore interesting behavior.

Figure 5.6. Dependence of short term standard deviation, σw, and

long term standard deviation, σtot, on gauge block length.

83

Table 5.2

Table of σwithin and σtotal by groups, in nanometers

Group σwithin σtotal

2 0.100 in. to 0.107 in. 5 43 0.108 in. to 0.126 in. 5 6 4 0.127 in. to 0.146 in. 5 5 5 0.147 in. to 0.500 in. 5 5 6 0.550 in. to 2.00 in. 6 8 7 3 in. 11 18

4 in. 15 41 5 in. 10 18 6 in. 12 48 7 in. 11 26 8 in. 10 76 10 in. 10 43 12 in. 10 25 16 in. 17 66 20 in. 13 29

Group σwithin σtotal 14 1.00 mm to 1.09 mm 5 615 1.10 mm to 1.29 mm 5 516 1.30 mm to 1.49 mm 5 517 1.50 mm to 2.09 mm 5 518 2.10 mm to 2.29 mm 5 519 2.30 mm to 2.49 mm 5 520 2.50 mm to 10 mm 5 521 10.5 mm to 20 mm 5 722 20.5 mm to 50 mm 6 823 60 mm to 100 mm 8 1824 125 mm 9 19

150 mm 11 38175 mm 10 22200 mm 14 37250 mm 11 48300 mm 7 64400 mm 11 58500 mm 9 56

92

The apparent lack of a strong length dependence in short term variability, as measured by σw, is theresult of extra precautions taken to protect the thermal integrity of longer blocks. Since the majorcause of variation for longer blocks is thermal instability the precautions effectively reduce the lengthdependence.

One notable feature of the above table is that short term variability measured as σw and long termvariability, measured as σtot, are identical within the measurement uncertainties until the size getslarger than 50 mm (2 in). For short sizes, this implies that long term variations due to comparators,environmental control and operator variability are very small.

A large difference between long and short term variability, as exhibited by the long sizes, can betaken as a signal that there are unresolved systematic differences in the state of the equipment or theskill levels of the operators. Our software system records the identity of both the operator and thecomparator for each calibration for use in analyzing such problems. We find that the differencesbetween operators and instruments are negligible. We are left with long term variations of thethermal state of the blocks as the cause of the larger long term variability. Since our current levelof accuracy appears adequate at this time and our thermal preparations, as described earlier, arealready numerous and time consuming we have decided not to pursue these effects.

5.7.1 Summary

Transferring length from master blocks to customer blocks always involves an uncertainty whichdepends primarily on comparator repeatability and the number of comparisons, and the accuracy ofthe correction factors used.

The random component of uncertainty (σtot) ranges from 5 nm (0.2 µin) for blocks under 25 mmto about 75 nm (3 µin) for 500 mm blocks. This uncertainty could be reduced by adding morecomparisons, but we have decided that the gain would be economically unjustified at this time.

Under our current practices no correction factors are needed for steel and chrome carbide blocks. For other materials a small added uncertainty based on our experience with correction factors isused. At this time the only materials other than steel and chrome carbide which we calibrate aretungsten carbide, chrome plated steel and ceramic, and these occur in very small numbers.

93

6. Gauge Block Interferometry

6.1 Introduction

Gauge Block calibration at NIST depends on interferometric measurements where the unit oflength is transferred from its definition in terms of the speed of light to a set of master gaugeblocks which are then used for the intercomparison process.

Gauge blocks have been measured by interferometry for nearly 100 years and the only majorchange in gauge block interferometry since the 1950's has been the development of the stabilizedlaser as a light source. Our measurement process employs standard techniques of gauge blockinterferometry coupled with an analysis program designed to reveal random and systematicerrors. The analysis program fosters refinements aimed at reducing these errors. A practicallimit is eventually reached in making refinements, but the analysis program is continued as ameasurement assurance program to monitor measurement process reliability.

Briefly, static interferometry is employed to compare NIST master gauge blocks with acalibrated, stabilized laser wavelength. The blocks are wrung to a platen and mounted in aninterferometer maintained in a temperature controlled environment. The fringe pattern isphotographed and at the same moment those ambient conditions are measured which influenceblock length and wavelength. A block length computed from these data together with the date ofmeasurement is a single record in the history file for the block. Analysis of this history fileprovides an estimate of process precision (long term repeatability), a rate of change of lengthwith time, and an accepted value for the block length at any given time.

Gauge block length in this measurement process is defined as the perpendicular distance from agauging point on the top face of the block to a plane surface (platen) of identical material andfinish wrung to the bottom face. This definition has two advantages. First, specifying a platenidentical to the block in material and surface finish minimizes phase shift effects that may occurin interferometry. Second, it duplicates practical use where blocks of identical material andfinish (from the same set) are wrung together to produce a desired length. The defined length ofeach block includes a wringing layer, eliminating the need for a wringing layer correction whenblocks are wrung together.

The NIST master gauge blocks are commercially produced gauge blocks and possess no unusualqualities except that they have a history of calibration from frequent and systematic comparisonswith wavelengths of light.

6.2 Interferometers

Two types of interferometers are used at NIST for gauge block calibration. The oldest is aKösters type interferometer, and the newest, an NPL Gauge Block Interferometer. They date

94

from the late 1930's and 1950's respectively and are no longer made. Both were designed formultiple wavelength interferometry and are much more complicated than is necessary for singlewavelength laser interferometry. The differences are pointed out as we discuss the geometry andoperation of both interferometers.

6.2.1 The Kösters Type Interferometer

Figure 6.1 Kösters type gauge block interferometer.

The light beam from laser L1 passes through polarization isolator B and spatial filter B', and isdiverged by lens C. This divergence lens is needed because the interferometer was designed to usean extended light source (a single element discharge tube). The laser produces a well collimatedbeam which is very narrow, about 1 mm. The spatial filter and lens diverges the beam to the properdiameter so that lens D can focus the beam on the entrance slit in the same manner as the light froma discharge tube. The entrance slit (S1) is located at the principle focus of lenses D and F. Lens Fcollimates the expanded beam at a diameter of about 35 mm, large enough to encompass a gaugeblock and an area of the platen around it.

The extended collimated beam then enters a rotatable dispersion prism, W. This prism allows onewavelength to be selected from the beam, refracting the other colors at angles that will not produceinterference fringes. An atomic source, such as cadmium (L) and a removable mirror M, can be usedin place of the laser as a source of 4 to 6 different calibrated wavelengths. If only a laser is used theprism obviously has no function, and it could be replaced with a mirror.

The rest of the optics is a standard Michelson interferometer. The compensator plate, I, is necessary

95

when using atomic sources because the coherence length is generally only a few centimeters and theeffective light path in the reference and measurement arms must be nearly the same to see fringes. Any helium-neon laser will have a coherence length of many meters, so when a laser is used thecompensator plate is not necessary.

The light beam is divided at the beam splitter top surface, BS, into two beams of nearly the sameintensity. One beam (the measuring beam) continues through to the gauge block surface G and theplaten surface P. The other beam (reference beam) is directed through the compensating plate I tothe plane mirror R. The beams are reflected by the surfaces of the mirror, platen, and gauge block,recombined at the beam splitter, and focused at the exit aperture, S2. A camera or human eye canview the fringes through the exit aperture.

Interference in this instrument is most readily explained by assuming that an image of referencemirror R is formed at R' by the beam splitter. A small angle, controlled by adjustment screws ontable T, between the image and the gauge block platen combination creates a Fizeau fringe pattern. When this wedge angle is properly adjusted for reading a gauge block length, the interferencepattern will appear as in figure 6.2.

Fig. 6.2 Observed fringe pattern for a gauge block wrung to a platen.

The table T has its center offset from the optical axis and is rotatable by remote control. Severalblocks can be wrung to individual small platens or a single large platen and placed on the table tobe moved, one at a time, into the interferometer axis for measurement.

Two criteria must be met to minimize systematic errors originating in the interferometer: (1) theoptical components must be precisely aligned and rigidly mounted and (2) the optical componentsmust be of high quality, i.e., the reference mirror, beam splitter and compensator must be plane andthe lenses free of aberrations.

96

Alignment is attained when the entrance aperture, exit aperture and laser beam are on a common axisnormal to, and centered on, the reference mirror. This is accomplished with a Gaussian eyepiece,having vertical and horizontal center-intersecting cross-hairs, temporarily mounted in place of theexit aperture, and vertical and horizontal center-intersecting cross-hairs permanently mounted on thereference mirror. Through a process of autocollimation with illumination at both entrance apertureand Gaussian eyepiece, the reference mirror is set perpendicular to an axis through the intersectionsof the two sets of cross-hairs and the entrance, and exit apertures are set coincident with this axis.

In addition, the laser beam is aligned coincident with the axis and the prism adjusted so the laserlight coming through the entrance aperture is aligned precisely with the exit aperture. Having anexact 90º angle between the measuring leg and the reference leg is not essential as can be seen frominstrument geometry. This angle is governed by the beam splitter mounting. All adjustments arechecked regularly.

The gauge block and its platen are easily aligned by autocollimation at the time of measurement andno fringes are seen until this is done. Final adjustment is made while observing the fringe pattern,producing the configuration of figure 6.2.

Temperature stability of the interferometer and especially of the gauge block being measured isimportant to precise measurement. For this reason an insulated housing encloses the entire Kostersinterferometer to reduce the effects of normal cyclic laboratory air temperature changes, radiated heatfrom associated equipment, operator, and other sources. A box of 2 cm plywood with a hingedaccess door forms a rigid structure which is lined with 2.5 cm thick foam plastic. Reflectivealuminum foil covers the housing both inside and out.

6.2.2 The NPL Interferometer

The NPL interferometer is shown schematically in figure 6.3. The system is similar in principle tothe Kösters type interferometer, although the geometry is rather different.

97

Figure 6.3 Schematic of the NPL interferometer.

The NIST version differs slightly from the original interferometer in the addition of a stabilized laser,L1. The laser beam is sent through a spatial filter B (consisting of a rotating ground glass plate) todestroy the temporal coherence of the beam and hence reduce the laser speckle. It is then sentthrough a diverging lens C to emulate the divergent nature of the atomic source L2. The atomicsource, usually cadmium, is used for multicolor interferometry.

The beam is focused onto slit, S1, which is at the focal plane of the D. Note that the entrance slit andexit slit are separated, and neither is at the focal point of the lens. Thus, the light path is notperpendicular to the platen. This is the major practical difference between the Kösters and NPLinterferometers. The Kösters interferometer slit is on the optical axis that eliminates the need for anobliquity correction discussed in the next section. The NPL interferometer, with the slit off theoptical axis does have an obliquity correction. By clever design of the slits and reflecting prism theobliquity is small, only a few parts in 106.

The beam diverges after passing through slit S1 and is collimated by lens F. It then passes throughthe wavelength selector (prism) W and down onto the Fizeau type interferometer formed by flat Rand platen P. The height of the flat can be adjusted to accommodate blocks up to 100 mm. Theadjustable mount that holds the flat can be tilted to obtain proper fringe spacing and orientationshown in figure 6.2.

The light passes back through the prism to a second slit S2, behind which is a small reflecting prism

OPTICAL FLAT

SEMI-REFLECTING COATING OF BISMUTH OXIDE

GAGE BLOCK

BASE PLATE

DISPERSION PRISMLENS

SLIT

REFLECTING PRISM

MERCURY 196

LIGHT SOURCE

CADMIUM

98

(RP) to redirect the light to an eyepiece for viewing.

The optical path beyond the slit is inside a metal enclosure with a large door to allow a gauge blockplaten with gauge blocks wrung down to be inserted. There are no heat sources inside the enclosureand all adjustments can be made with knobs outside the enclosure. This allows a fairlyhomogeneous thermal environment for the platen and blocks.

Multicolor interferometry is seldom needed, even for customer blocks. A single wavelength isadequate if the gauge block length is known better than 1/2 fringe, about 0.15 µm (6 µin). Forcustomer calibrations, blocks are first measured by mechanical comparison. The length of the blockis then known to better than 0.05 µm (2 µin), much better than is necessary for single wavelengthinterferometry.

Since the NPL interferometer is limited to relatively short blocks the need for thermal isolation isreduced. The temperature is monitored by a thermistor attached to the gauge block platen. Onethermometer has proven adequate and no additional insulation has been needed.

6.2.3 Testing Optical Quality of Interferometers

Distortion caused by the system optics, is tested by evaluating the fringe pattern produced on amaster optical flat mounted on the gauge block platen support plate. Photographs showing fringesoriented vertically and horizontally are shown in figure 6.4. Measurements of the fringe patternindicate a total distortion of 0.1 fringe in the field. A correction could be applied to each blockdepending on the section of the field used, but it would be relatively small because the field sectionused in fringe fraction measurement is small. Generally the corrections for optical distortion in bothinstruments are too small to matter.

Figure 6.4 The quality of the optics of the interferometer can be measured byobservations of the fringes formed by a calibrated reference surface such as an opticalflat or gauge block platen.

99

6.2.4 Interferometer Corrections

In some interferometers, the entrance and exit apertures are off the optical axis so light fallsobliquely on the gauge block. The NPL interferometer is one such design.

In the Kosters type interferometer these apertures are aligned on the optical axis and consequentlythe illumination is perpendicular to the gauge block and platen surfaces. In figure 6.5 the case fornormal incidence is given. Note that the light reflected from the surfaces follows the same pathbefore and after reflection. At each point the intensity of the light depends on the difference in phasebetween the light reflected from the top surface and that reflected from the lower surface. Wheneverthis phase difference is 180º there will be destructive interference between the light from the twosurfaces and a dark area will be seen. The total phase difference has two components. The firstcomponent results from the electromagnetic phase difference of reflections between the glass to airinterface (top flat) and the air to glass interface (bottom flat). This phase shift is (for non-conductingplatens) 180 º. The second phase difference is caused by the extra distance traveled by the lightreflected at the bottom surface. The linear distance is twice the distance between the surfaces, andthe phase difference is this difference divided by the wavelength, λ.

Figure 6.5 In the ideal case of normal incidence , a dark fringe will appear wherever the totalpath difference between the gauge block and platen surfaces is a multiple of the wavelength.

At every position of a dark fringe this total phase difference due to the path difference is a multipleof λ:

∆(path) = nλ = 2D (6.1)

100

where n is an integer.

If the wedge angle is α, then the position on the nth fringe is

X = (nλ)/(2tan(α)) (6.2)

For the second case, where the light is not normal to the block and platen, we must examine the extradistance traveled by the light reflecting from the lower flat and apply a compensating factor calledan obliquity correction. Figure 6.6 shows this case. The difference in path between the lightreflected from the upper reference surface and the lower reference surface must be a multiple of λto provide a dark fringe. For the nth dark fringe the distance 2L must equal nλ.

Figure 6.6 When the light is not normal to the surface the fringes areslightly displaced by an amount proportional to the distance betweenthe surfaces.

The equation for a dark fringe then becomes

∆(path) = nλ = 2L (6.3)

From the diagram we see that L is:

L = D cos(θ) = nλ (6.4)

DL L

D'

fringe shift

101

Thus the new perpendicular distance, D', between the two flats must be larger than D. This impliesthat the nth fringe moves away from the apex of the wedge, and the distance moved is proportionalto cos(θ) and distance D.

When a gauge block is being measured, the distance from the top flat to the bottom flat is larger thanthe distance from the top flat to the gauge block surface. Thus the fringes on the block do not shiftas far as the fringes on the flat. This larger shift for the reference fringes causes the fringe fractionto be reduced by a factor proportional the height of the block, H, and cos(θ). Note that this effectcauses the block to appear shorter than it really is. Since the angle θ is small the cosine can beapproximated by the first terms in its expansion

(6.5)

and the correction becomes:

Obliquity Correction ~ (H/2) θ2 (6.6)

This correction is important for the NPL Hilger interferometer where the entrance slit andobservation slit are separated. In the Kösters type interferometer the apertures are precisely on theoptical axis, the obliquity angle is zero, and thus no correction is needed.

In most interferometers, the entrance aperture is of finite size, therefore ideal collimation does notoccur. Light not exactly on the optical axis causes small obliquity effects proportional to the sizeof the aperture area. A laser, as used with these interferometers, is almost a point source because theaperture is at the common focal point of lens D, and collimating lens E. At this point the beamdiameter is the effective aperture. When using diffuse sources, such as discharge lamps, the effectiveaperture is the physical size of the aperture.

The theory of corrections for common entrance aperture geometries has been worked out in detail[37,38,39]. The correction factor for a gauge block of length L, measured with wavelength λ, usinga circular aperture is approximately [40]

(6.7)

where L is the block length, D is the aperture diameter, and f is the focal length of the collimatinglens.

+ 24

+ 2

- 1 =)( 42

cos

f16DLx = 2

2(Circle) CorrectionSlit

102

For rectangular apertures the approximate correction is

(6.8)

where L is the block length, h and l the height and length of the slit, and f the focal length of thecollimating lens.

As aperture diameter approaches zero, the correction approaches zero. The correction is very smallwhen using a laser because the beam diameter is small at the aperture.

6.2.5 Laser Light Sources

The advantage of a laser light source lies in its unequalled coherence. Conventional spectral lightsources have such low coherence that blocks longer than 300 mm have, in the past, been measuredby stepping up from shorter blocks. Laser fringe patterns in these interferometers compared withpatterns from other spectral lamps are of superior contrast and definition through the length range0 to 500 mm.

The disadvantage stems from the somewhat unstable wavelength of lasers. This problem has beenovercome by a number of different stabilization schemes, and by using the .6328 µm line of HeNelasers. These lasers can be related to the definition of the meter by calibration against an iodinestabilized HeNe laser [41,42]. The wavelength of the iodine stabilized laser is known to a few partsin 109 and is one of the recommended sources for realizing the meter [8]. The lasers currently usedfor gauge block calibration are polarization stabilized [43]. Single wavelength static interferometry, in contrast with multiple wavelength interferometry,requires that the gauge block length be known to within 0.25 wavelength (0.5 fringe) either from itshistory or from another measurement process. This is no problem for NIST reference blocks, or forblocks calibrated by the NIST mechanical comparison technique.

Laser light must not be allowed to reflect back into its own cavity from the interferometer opticalcomponents because this will disturb the lasing action and may cause a wavelength shift, wavelengthjittering or it may stop the lasing altogether. The reflected light is prevented from re-entering thelaser by a polarization isolator consisting of a Glan Thompson prism and a quarter wave plate in thebeam as it emerges from the cavity. This assembly is tilted just enough to deflect reflections fromits faces to the side of the laser exit aperture.

6.3 Environmental Conditions and their Measurement

Environmental control of the laboratory holds temperature at 20.05 ºC and water vapor content

f24

)l+hLx(= 2

22)(RectangleCorrectionSlit

103

below 50% relative humidity. Temperature variations within the insulated interferometer housingare attenuated by a factor of about 10 from those in the room, thus insuring stability of bothinterferometer and blocks.

Temperature, atmospheric pressure, and water vapor content of the ambient air in the interferometerlight path must be measured at the time the gauge block is measured. From these properties therefractive index of the air is calculated, and in turn, the laser wavelength. Since interferometry is acomparison of the light wavelength to the block length the accuracy of the wavelength is of primaryimportance. The gauge block temperature is measured so that the block length at 20 ºC, the standardtemperature for dimensional metrology, can be computed.

6.3.1 Temperature

The measuring system accuracy for air and block temperature depends on the length of the gaugeblock to be measured. Since both the wavelength and block temperature corrections are lengthdependent, a 500 mm measurement must have correction factors 10 times as good as those for a 50mm measurement to achieve the same calibration accuracy,

For blocks measured with the NPL interferometer (less than 100 mm) a calibrated thermistor probeis used to measure temperature to an accuracy of 0.01 ºC and a calibrated aneroid barometer is usedto measure atmospheric pressure to 0.1 mm of mercury.For long blocks the system is much more complicated. The measuring system for air and blocktemperature consists of copper-constantan thermocouples referenced to a platinum resistancethermometer. Figure 6.7 is a schematic diagram of the system.

A solid copper cube provides a stable reference temperature measured by a Standard PlatinumResistance Thermometer (SPRT) and a Mueller resistance bridge. A well in the cube holds thethermometer stem, the reference thermocouple junction, and a liquid of high heat conductivity butlow electrical conductivity. There is also a second well whose function will be described later. Anenclosed thermocouple selector switch connects the measuring junctions, one at a time, to thereference junction and the thermal emfs generated by the temperature differences between measuringjunction and reference junction are indicated on a nanovoltmeter. The nanovoltmeter is shared withthe bridge where it is used as a nullmeter. To keep the copper block in a stable state and to minimizethermal emfs in the selector switch, both block and switch are enclosed in an insulated box. Theswitch is operated through a fiber shaft extending to the exterior and the protruding part of thethermometer is protected by an insulated tube.

104

Figure 6.7 The temperature measurement system used for gauge block calibrationconsists of multiple thermocouples referenced to a calibrated platinum resistancethermometer.

Thermocouples generate emfs proportional to the temperature gradient along the wire:

(6.9)

where T1 and T2 are the temperatures of the reference and measuring junctions respectively, and φis the thermocouple constant. Minimizing this gradient reduces uncertainties.

Relating the system to the International Temperature Scale of 1990 (ITS '90) is a three stepprocedure. First, the SPRT is calibrated using methods described in NIST Technical Note 1265 [44]. Second, the bridge is calibrated as described the same monograph. The third step is the calibrationof the thermocouples, which is accomplished with a second SPRT and insulated copper cube. Thesecond cube has an extra well for the thermocouples being calibrated and is heated one or twodegrees to equilibrium. Measured thermal emfs generated by the temperature difference betweenthe two cubes together with the temperatures of the cubes allow computation of a calibration factorfor each junction.

Integrity is maintained by periodic check of the SPRT with a triple point cell, checking the bridgeagainst a calibrated resistor and checking the thermocouples for equality when they are all at thesame temperature in a well of an insulated copper cube.

6.3.2 Atmospheric Pressure

The interferometer housing is not airtight and pressure inside it is the same as laboratory pressure.

PLATINUM RESISTANCETHERMOMETER

REFERENCETHERMOCOUPLE

JUNCTION

MEASURING

THERMOCOUPLEJUNCTIONS

DPDT SWITCH

RESISTANCE

THERMOMETERBRIDGE

NANO-

VOLTMETER REFERENCETEMPERATURE

THERMOCOUPLE

SELECTOR

SWITCH

dT = E TT

12

105

An aneroid barometer is located adjacent to and at the same elevation as the interferometer. Theaneroid used for gauge block measurements has good stability, and frequent comparisons with NISTreference barometers insure its reliability.

For long blocks more precise measurement of the pressure is useful. There are a number ofcommercially available atmospheric pressure sensors available which are of adequate accuracy forgauge block calibration. Remembering that measurements to one part in 107 requires pressuremeasurement accuracy of .3mm of Hg, a calibrated aneroid barometer can be adequate. For systemsrequiring automatic reading of pressure by a computer there are a number of barometers which usecapacitance or vibrating cylinders to measure pressure and have RS-232 or IEEE 488 interfaces.

6.3.3 Water Vapor

The humidity is derived from a chilled mirror dew point hygrometer. The system is calibrated bythe NIST humidity calibration service and is accurate to about 1% R.H. Other systems haveadequate accuracy but need more frequent checking for drift. The disadvantage of the chilled mirroris that it requires a specified air flow rate not usually available inside an interferometer. Humidityinside an interferometer enclosure may differ from the laboratory value by as much as 6% R.H.,especially during periods of changing laboratory humidity. If this presents a serious problem (seetable 6.1) then a hygrometer type less dependent on air flow rate must be mounted inside theinterferometer.

6.4 Gauge Block Measurement Procedure

In preparation for measurement, the gauging faces of the NIST standards are cleaned with ethylalcohol. Wiping the alcohol-rinsed gauging faces is done with paper tissue or a clean cotton towel. Lint or dust is removed with a camel's hair brush.

The gauging faces are examined for burrs and, if present, these are removed with a deburring stone. The bottom end is tested for wringability with a quartz optical flat. The transparency of the flatmakes possible a judgment of the wring. A uniform grey color at the interface indicates a goodwring. Any colored or light areas indicate a gap of more than 25 nm between the faces in that area,and such a gap may result in an erroneous length measurement. Deburring and cleaning is continueduntil a good wring is achieved. The quartz flat is left wrung to the gauging face to keep it clean untilpreparations are completed for wringing the block to the steel optical flat (platen).

Preparation of the steel platen is similar. Ethyl alcohol is used to rinse the face, then deburring ifnecessary, followed by more rinsing and drying. After sweeping with a camel hair brush, a thinuniform film of light grease is applied to the wringing surface with a bit of tissue. This film isrubbed and finally polished with tissue or towel until the film appears to be gone. After carefullysweeping the platen with a brush, it is ready for wringing.

The block is removed from the quartz flat and the exposed gauging face is immediately wrung by

106

carefully sliding it onto the edge of the platen, maintaining perpendicularity by feel. Sliding iscontinued until wringing resistance is felt, and then with slight downward pressure the block isslowly worked into its final position. Square style blocks, such as the NIST long blocks, arepositioned so that the gauge point is at the right in the viewing field of the interferometer. One tofour blocks can be wrung to the platens used in the Kosters interferometer. Ten square blocks or 20rectangular blocks can be wrung onto the platens at one time for use in the NPL interferometer.

For long blocks measured in the Kosters interferometer, the platen with its wrung blocks is placedin the interferometer and two thermocouples are fed down the center hole of each block, one aboutthree quarters down and one about one quarter down the length of the block. A small wad ofpolyurethane is pushed into the hole to seal and hold the wires. For small blocks, where there canbe no significant vertical thermal gradient, the temperature sensor is attached to the platen.

A preliminary alignment by autocollimation with the Gaussian eyepiece as the support table isadjusted, will produce fringes. Rapid changes in the fringe pattern occur as the blocks and platencool from handling.

It is convenient to wring the blocks in late afternoon and allow overnight temperature stabilization. Two length observations are made several hours apart, but the first measurement is not taken untilthe laser is in equilibrium. Leaving the laser on continuously when observations are being made overseveral days eliminates delays.

Final fringe pattern adjustment is made so that the platen fringes are parallel to the top and bottomedges of the block and one fringe on the block goes through its defined gauge point. For the Kostersinterferometer the direction of increasing fringe order is verified by pressing down on the eyepieceand observing the fringe movement. In the NPL, the reference mirror is pushed and the fringemovement is observed. The fringe fraction can be estimated by eye or a photograph can be takenof the pattern. Block temperature, air temperature, barometric pressure, and humidity are measuredimmediately before and after the observation and then averaged.

The photograph can be either a positive or a negative. A compromise is made between image sizeand exposure time, but short exposure is desirable because the fringe pattern may shift during a longexposure. Further image magnification takes place in the fringe measuring instrument. Acompromise is also made between magnification in the photograph and magnification in theinstrument used to measure the photograph. Too much total magnification degrades imagesharpness, making fringe measurements more difficult.

Fringe fractions are obtained from the photograph with a coordinate comparator having a fixed stagein the focal plane of a microscope movable on X-Y slides by precision screws with drum readouts. Four measurements of distance "a" and "b" are recorded and averaged to obtain fringe fraction f, asshown in figure 6.8.

107

Figure 6.8. The fringe fraction is the fractional offset of the fringes on the blockfrom those of the platen.

Settings on the block fringe to obtain "a" are made at the point where the fringe intersects the gaugepoint. Settings on the platen fringes to obtain "a" and "b" are made as close as practical to the edgesof the block because it is here that the platen best represents an extension of the bottom face of theblock. Parallelism between gauge block and platen is also read from the photo because thisinformation is useful in detecting a defective wring. Poor wringing contact is indicated if theparallelism is different from its usual value (or mechanically determined value using the comparator)and this is sufficient cause to discard a particular length measurement.

Photographing fringe patterns has several advantages. The photograph is a permanent record thatcan be reinterpreted at any time with the same or other techniques. Changes in block geometry arereadily seen by comparing photographs in time sequence. The coordinate comparator is superior tothe common practice of fringe fraction estimation by eye because it is more objective, lends itselfto multiple measurements and averaging, and is more precise. Finally, photography is fast and thuspermits readings of the ambient conditions to closely bracket the fringe recording. This is especiallyimportant because atmospheric pressure is generally not constant.

The advent of computer vision systems provides another method of recording and analyzing fringes. There are commercially available software packages for measuring fringe fractions that can beincorporated into an automated gauge block measuring system.

6.5 Computation of Gauge Block Length Calculating gauge block length from the data is done in 3 steps:

108

1. Calculation of wavelength, λtpf, at observed ambient conditions.

2. Calculation of the whole number of fringes in the gauge block length from its predictedlength and the laser wavelength in ambient air.

3. Calculation of the gauge block length from the whole number of fringes, the observedfraction, wavelength in ambient air, gauge block temperature, and interferometric correctionfactors.

6.5.1 Calculation of the Wavelength

Edlen's 1966 formula [45] for the refractive index of air is used to calculate the wavelength. Comparison of absolute refractometers showed that the Edlen equation agreed with refractometersas well as the refractometers agreed with each other, to about 1 part in 107 [46]. More recent workby Birch and Downs [47] has shown that the original data used for the water vapor correction wasslightly in error. The new correction (the parameter C below) is thought to increase the accuracy ofthe Edlen formula to about 3x10-8 . This latter figure can be taken as an estimate of the accuracy ofthe formula.

The latest revision of the Edlen formula is [48]:

λ0 = ntpfλtpf (6.10)

and thus λ0

λtpf = ----------- = λ0 [1 + A*B - C]-1 (6.11)

ntpf

where

A = p[8342.22 + 2406057(130-σ2)-1 + 15997 (38.9-σ2)-1] 10-8 (6.12)96095.43

B = 1 + p (0.601-0.00972t) 10-6 (6.13)1+0.003661t

C = f (3.7345-0.0401σ2) 10-8. (6.14)

109

The symbols used are:

λ0 = Vacuum wavelength in µm

λptf = Wavelength at observed conditions t,p,f

nptf = Index of refraction of air at t,p,fp = Atmospheric pressure in Pat = Temperature in degrees Celsiusf = water vapor pressure in Paσ = 1/λ0 in µm.

6.5.2 Calculation of the Whole Number of Fringes

This method for determining the whole number of fringes is valid if the block length is known tobetter than 0.5 fringe ( 1 fringe is about 0.32 µm for helium neon laser light) from either its historyor from an independent measurement process. The calculation is as follows:

2Lp[1+C(t'-20)]F = -------------------- (6.15)

λtpf

where F is the number of interference fringes in the length of the block at temperature t', andin air of temperature t, pressure p, and vapor pressure f.

Lp is the predicted block length at 20 ºC taken from its history or an independentmeasurement.

C is the linear thermal expansion coefficient per degree Celsius of the block.

t' is the block temperature at the time of measurement.

λtpf is the wavelength in the ambient air at the time of measurement.

The fractional part of F is retained temporarily for the reasons explained below.

110

6.5.3 Calculation of Block Length from the Observed Data

Generally, the observed fringe fraction φ is simply substituted for the fractional part of F, but thereare cases where the last digit in whole number F must be raised or lowered by one. For example ifthe fractional part of F is 0.98 and the observed fraction is 0.03, obviously, the last whole numberin F must be raised by one before substituting the observed fraction.

The new measured block value, in fringes, is

F' = F + φ. (6.16)

Finally, the interferometer aperture correction, δ2, is added, the block is normalized to 20 ºC, and aconversion to length units is made to arrive at the final value at 20 ºC as follows:

λtpfL20 = ----- F'[1+C(20-t')] + δ2 . (6.17)

2

6.6 Type A and B Errors

Potential sources of errors are as follows:

1. Wavelengtha. Vacuum wavelength of laserb. Refractive index of air equationc. Refractive index determination

Air temperature measurementAtmospheric pressure measurementHumidity measurement

2. Interferometera. Alignmentb. Aperture correctionc. Obliquity correction

3. Gauge Blocka. Gauge block temperature measurementb. Thermal expansion coefficientc. Phase shift difference between block and platen

111

Length proportional errors are potentially the most serious in long block measurement. Their relativeinfluence is illustrated in table 10. One example will help explain the table: if air temperaturemeasurement is in error by 0.1 ºC a systematic error of 1 part in 107 results. For a 250 mm gaugeblock the error is 25 nm, a significant error. On a 10 mm gauge block the same temperaturemeasurement leads to only a 1 nm error, which is negligible.

In our measurement process every practical means, as previously discussed, was used to reduce theseerrors to a minimum and, at worst, errors in measuring the parameters listed above do not exceedthose listed in table 10. Further evaluation of these errors is described in section 4.3 on processcontrol and our master block history.

A few sources of error in the list of systematic errors but not in the table have not been completelydiscussed. The first of these is phase shift difference between block and platen. When light isreflected from a surface the phase of the reflected light is shifted from the phase of the incident bysome amount which is characteristic of the surface's material composition and surface finish. Theexplanation and measurement of phase is given in appendix C. Most gauge blocks have phase shiftsthat differ from quartz by an equivalent of 25 to 75 nm. This is a large source of error, although ifboth the block and platen are made of the same metal the error is reduced below 20 nm. Since thephase shift occurs only at the metal surface it is not length dependent.

Table 10

Effects of Environmental Factor Errors on Length

Parameter Change Which Leads to a Length Error of 1 part in 107

WavelengthVacuum wavelength 1 part in 107 Refractive index formula 1 part in 107

Air temperature 0.1ºC Atmospheric Pressure 40 paRelative Humidity 12% R.H.

Interferometer Alignment 0.45 mm/mGauge Block

Steel gauge block temperature 0.009 ºC Chrome Carbide block temp. 0.012 degc

Thermal expansion coefficients of gauge blocks have an uncertainty that is, to a large extent,nullified by measuring at temperatures very close to the 20 ºC reporting temperature. Gauge block

112

manufacturers give 11.5 x 10-6/ºC for steel blocks. This average value may be uncertain by up to0.6 x 10-6/ºC for individual blocks. The expansion coefficient for steel blocks is also lengthdependent for long blocks, as discussed earlier in section 3.3. The expansion coefficient for longersizes can differ from the nominal 11.5 x 10-6/ºC by nearly 1 x 10-6/ºC. These uncertainties can causesignificant errors if blocks are used at other than 20 ºC.

There are, of course, random errors that arise from variability in any parameter in the table, and inaddition, they can arise in the fringe fraction measurement and wringing variability. Wringing isdiscussed in more detail in appendix B. Gauge blocks are generally measured for 3 to 5 separatewrings to sample variability due to these sources.

6.7 Process Evaluation

Figure 6.9 shows some examples of the interferometric history of NIST master gauge blocks. Asexpected, the scatter in the data is larger for longer blocks because of the uncertainties in the lengthdependent thermal expansion and index of refraction corrections.

Figure 6.9. Examples of the interferometric history of NIST master gauge blocks.

ID: F15 Size: 25 mm 15 Pts. Avg. = 0.018

Nom

inal

Dev

iatio

n (m

icro

met

ers)

ID: 5 83 Size: 0.1 inch 56 Pts. Avg. = 0.296

Deviation in m

icrometersD

evia

tion

in m

icro

inch

es

113

The length dependence of the variability is shown graphically in figure 6.10. This variability hastwo components: (1) non-length related variabilitys due to variations in the wringing film betweenthe block and the platen, and random errors in reading the fringes; and (2) a length dependent partprimarily due to environmental effects.

Figure 6.10. The variability (σ) of NIST master gauge blocks as a function of length.

The major components of the interferometric measurement uncertainty are shown in table 6.2. Theerrors can be divided into two classes: those which are correlated ( the same for all measurements)and those which are uncorrelated (vary for each measurement). The classification depends on themeasurement history. NIST gauge blocks which are measured over a long period of time, duringwhich the thermometers and barometers are recalibrated. In this case the errors due to thermalexpansion and refractive index corrections vary. For a customer block, measured 4 times over a 4day period, they do not vary. Table 6.2 classifies the errors for both NIST masters and customercalibrations.

Mic

roin

ches N

anometers

Nominal Length (inches)

114

Table 6.2

Uncertainty UncertaintySource of Uncertainty Type for NIST Masters Type for Customers

Reading error uncorrelated uncorrelated

Obliquity/Slit Correction correlated correlated

Wringing Film Thickness uncorrelated uncorrelated

Phase Correction correlated correlated

Thermal Expansion Correction Temperature Measurement uncorrelated correlated Expansion Coefficient Value correlated correlated

Index of Refraction Edlén Equation correlated correlated Environmental Measurements uncorrelated uncorrelated Laser Frequency uncorrelated correlated

The interferometric history provides a statistical sample of all of the uncorrelated errors. Using theinformation from figure 6.10 we obtain an estimate of the uncorrelated errors to be:

Uu (µm) ~ 0.008 + 0.03 x L (L in meters) (6.20)

Multiple measurements of a gauge block can reduce this figure; if n measurements are made theuncertainty of the average length becomes Uu/n. Unfortunately, the correlated errors are not sampledby repeated measurements and are not reduced by multiple measurements. An estimate of thecorrelated uncertainties for NIST master gauge block calibrations is:

Uc (µm) ~ 0.007 + 0.07 x L (L in meters) (6.21)

Because customer blocks have more correlated errors and more importantly do not have measuredthermal expansion coefficients, the use of multiple measurements is less effective in reducingmeasurement uncertainty. For most customers, interferometric measurements are not cost-effectivesince the uncertainty for a three wring interferometric measurement is not significantly lower thanthe mechanical comparison measurement, but costs twice as much. If, however, a customer sendsin the same blocks for interferometric measurement over a number of years and uses the accumulatedhistory then the uncertainty of the blocks will approach that of NIST master blocks.

115

6.8 Multiple wavelength interferometry

If we think of gauge block interferometry as the measurement of height with a ruler marked in unitsof λ/2, we quickly realize that our ruler has units unlike a normal ruler. The lines which show adistance of λ/2 are there but there are no numbers to tell us which mark we are looking at. This rulercan be used if we know the length of the block to better than λ/4 since we can use this distance andmake the small correction (less than λ/4) found by interpolating between the lines on our scale. Suppose, however, we do not know the length this well ahead of time.

The correct method in this case is to use a number of rulers with different and incommensuratescales. In interferometry we can use several light sources ranging from blue to red to provide thesedifferent scales. A simple example is shown in figure 6.11.

Figure 6.11. Four color interferometry is equivalent to measuringa length with four scales of known pitch which have indeterminatezeros.

For each scale (color) there will be a fringe fraction fi. With only one color the block might have any

f = 0 f = .71 f = .87 f = .62

/2 = 10 2/ = 17 2/ = 23 2/ = 29

116

length satisfying the formula:

L = (n1 + f1)*λ1/2 where n1 is any integer. (6.22)

If we look at a second color, there will be another fringe fraction f2. It also will be consistent withany block length satisfying the formula:

L = (n2 + f2)*λ2/2 where n2 is any integer. (6.23)

However, we have gained something, since the number of lengths which satisfy both relations isvery much reduced, and in fact are considerably further apart than λ/2. With more colors the numberof possible matches are further reduced until a knowledge of the length of the block to the nearestcentimeter or more is sufficient to determine the exact length.

In theory, of course, knowing the exact fringe fractions for two colors is sufficient to know anylength since the two wavelengths are not commensurate. In practice, our knowledge of the fringefractions is limited by the sensitivity and reproducibility of our equipment. In practice, 1/20 of afringe is a conservative estimate for the fringe fraction uncertainty. A complete analysis of theeffects of the uncertainty and choice of wavelengths on multicolor interferometry is given by Tilford[49].

Before the general availability of computers, the analysis of multicolor interferometry was a timeconsuming task [50]. There was a large effort made to produce calculation aids in the form of booksof fringe fractions for each popular source wavelength, correction tables for the index of refraction,and even fringe fraction coincidence rules built somewhat like a slide rule. Since the advent ofcomputers it is much easier to take the brute force approach since the calculations are quite simplefor the computer.

Figure 6.11 shows a graphical output of the NIST multicolor interferometry program using acadmium light source. The program calculates the actual wavelengths for each color using theenvironmental factors (air temperature, pressure and humidity). Then, using the observed fringefraction, shows the possible lengths of the gauge block which are near the nominal length for eachcolor. Note that the possible lengths are shown as small bars, with their width corresponding to theuncertainty in the fringe fraction.

117

Figure 6.12. Graphical output of the NIST multicolor interferometry program. The length at which all the colors match within 0.030 µm is assumed to be the true gauge block length. The match is at -0.09 µm in this example.

The length where all of the fringes overlap is the actual length of the block. If all the fringes do notoverlap in the set with the best fit, the inconsistency is taken as evidence of an operator error and theblock is re-measured. The computer is also programmed to examine the data and decide if there isa length reasonably close to the nominal length for which the different wavelengths agree to a giventolerance. As a rule of thumb, all of the wavelengths should agree to better than 0.030 µm to beacceptable.

Analytic methods for analyzing multicolor interferometry have also been developed [49]. Ourimplementations of these types of methods have not performed well. The problem is probably thatthe span of wavelengths available, being restricted to the visible, is not wide enough and the fringefraction measurement not precise enough for the algorithms to work unambiguously.

6.9 Use of the Linescale Interferometer for End Standard Calibration

There are a number of methods to calibrate a gauge block of completely unknown length. Themultiple wavelength interferometry of the previous section is used extensively, but has the limitationthat most atomic sources have very limited coherence lengths, usually under 25 mm. The methodcan be used by measuring a set of blocks against each other in a sequence to generate the longestlength. For example, for a 10 inch block, a 1 inch block can be measured absolutely followed bydifferential measurements of a 2 inch block with the 1 inch block, a 3 inch block with the 2, a 4 inchblock with the 3, a 5 inch block with the 4, and the 10 inch block with the 2, 3 and 5 inch blocks

0-.1-.2-.3 .1 .2 .3

Micrometers from Nominal

-.4 .4

LaserCd RedCd GreenCd BlueCd Violet

118

wrung together. Needless to say this method is tedious and involves the uncertainties of a largenumber of measurements.

Another, simpler, method is to convert a long end standard into a line standard and measure it withan instrument designed to measure or compare line scales (for example, meter bars) [51]. The NISTlinescale interferometer, shown schematically below, is generally used to check our master blocksover 250 mm long to assure that the length is known within the 1/2 fringe needed for singlewavelength interferometry.

The linescale interferometer consists of a 2 m long waybed which moves a scale up to a meter inlength, under a microscope. An automated photoelectric microscope, sends a servo signal to themachine controller which moves the scale so that the graduation is at a null position on themicroscope field of view. A laser interferometer measures the distances between the marks on thescale via a corner cube attached to one end of the scale support. This system is described in detailelsewhere [52,53].

To measure an end standard, two small gauge blocks that have linescale graduations on one side, arewrung to the ends of the end standard, as shown in figure 6.13. This "scale" is then measured on thelinescale interferometer. The gauge blocks are then removed from the end standard and wrungtogether, forming a short scale. This "scale" is also measured on the interferometer. The differencein length between the two measurements is the physical distance between the end faces of the endstandard plus one wringing film. This distance is the defined length of the end standard.

Figure 6.13. Two small gauge blocks with linescale graduations onone side are wrung to the ends of the end standard, allowing the endstandard to be measured as a linescale.

119

The only significant problem with this method is that it is not a platen to gauge point measurementlike a normal interferometric measurement. If the end standard faces are not flat and parallel themeasurement will not give the exact same length, although knowledge of the parallelism and flatnesswill allow corrections to be made. Since the method is only used to determine the length within 1/2fringe of the true length this correction is seldom needed.

120

7. References

[1] C.G. Peters and H.S. Boyd, "Interference methods for standardizing and testing precisiongauge blocks," Scientific Papers of the Bureau of Standards, Vol. 17, p.691 (1922).

[2] Beers, J.S. "A Gauge Block Measurement Process Using Single Wavelength Interferometry,"NBS Monograph 152, 1975.

[3] Tucker, C.D. "Preparations for Gauge Block Comparison Measurements," NBSIR 74-523.

[4] Beers, J.S. and C.D. Tucker. "Intercomparison Procedures for Gauge Blocks UsingElectromechanical Comparators," NBSIR 76-979.

[5] Cameron, J.M. and G.E Hailes. "Designs for the Calibration of Small Groups of Standardsin the Presence of Drift," NBS Technical Note 844, 1974.

[6] Klein, Herbert A., The Science of Measurement, Dover Publications, 1988.

[7] Galyer, J.F.W. and C.R. Shotbolt, Metrology for Engineers, Cassel & Company, Ltd.,London, 1964.

[8] "Documents Concerning the New Definition of the Meter," Metrologia, Vol. 19, 1984.

[9] "Use of the International Inch for Reporting Lengths of Gauge Blocks," National Bureau ofStandards (U.S.) Letter Circular LC-1033, May, 1959.

[10] T.K.W. Althin, C.E. Johansson, 1864-1943, Stockolm, 1948.

[11] Cochrane, Rexmond C., AMeasures for Progress," National Bureau of Standards (U.S.),1966.

[12] Federal Specification: Gauge Blocks and Accessories (Inch and Metric), FederalSpecification GGG-G-15C, March 20, 1975.

[13] Precision Gauge Blocks for Length Measurement (Through 20 in. and 500 mm),ANSI/ASME B89.1.9M-1984, The American Society of Mechanical Engineers, 1984.

[14] International Standard 3650, Gauge Blocks, First Edition, 1978-07-15, 1978..

[15] DIN 861, part 1, Gauge Blocks: Concepts, requirements, testing, January 1983.

[16] M.R. Meyerson, T.R. Young and W.R. Ney, "Gauge Blocks of Superior Stability: InitialDevelopments in Materials and Measurement," J. of Research of the National Bureau ofStandards, Vol. 64C, No. 3, 1960.

121

[17] Meyerson, M.R., P.M. Giles and P.F. Newfeld, "Dimensional Stability of Gauge BlockMaterials," J. of Materials, Vol. 3, No. 4, 1968.

[18] Birch, K.P., "An automatic absolute interferometric dilatometer," J. Phys. E: Sci. Instrum,Vol. 20, 1987.

[19] J.W. Berthold, S.F. Jacobs and M.A. Norton, "Dimensional Stability of Fused Silica, Invar,and Several Ultra-low Thermal Expansion Materials," Metrologia, Vol. 13, pp. 9-16 (1977).

[20] C.W. Marshall and R.E. Maringer, Dimensional Instability, An Introduction, Pergamon Press,New York (1977).

[21] Hertz, H., "On the contact of elastic solids," English translation in Miscellaneous Papers, Macmillan, N.Y., 1896.

[22] Poole, S.P., "New method of measuring the diameters of balls to a high precision,"Machinery, Vol 101, 1961.

[23] Norden, Nelson B., "On the Compression of a Cylinder in Contact with a Plane Surface,"NBSIR 73-243, 1973.

[24] Puttock, M.J. and E.G. Thwaite, "Elastic Compression of Spheres and Cylinders at Point andLine Contact," National Standards Laboratory Technical Paper No. 25, CSIRO, 1969.

[25] Beers, John, and James E. Taylor, "Contact Deformation in Gauge Block Comparisons," NBSTechnical Note 962, 1978

[26] Beyer-Helms, F., H. Darnedde, and G. Exner. "Langenstabilitat bei Raumtemperatur vonProben er Glaskeramik 'Zerodur'," Metrologia Vol. 21, p49-57 (1985).

[27] Berthold, J.W. III, S.F. Jacobs, and M.A. Norton. "Dimensional Stability of Fused Silica,

Invar, and Several Ultra-low Thermal Expansion Materials," Metrologia, Vol. 13, p9-16(1977).

[28] Justice, B., "Precision Measurements of the Dimensional Stability of Four Mirror Materials,"Journal of Research of the National Bureau of Standards - A: Physics and Chemistry, Vol.79A, No. 4, 1975.

[29] Bruce, C.F., Duffy, R.M., Applied Optics Vol.9, p743-747 (1970).

[30] Average of 1,2,3 and 4 inch steel master gauge blocks at N.I.S.T.

[31] Doiron, T., Stoup, J., Chaconas, G. and Snoots, P. "stability paper, SPIE".

122

[32] Eisenhart, Churchill, "Realistic Evaluation of the Precision and Accuracy of InstrumentCalibration Systems," Journal of Research of the National Bureau of Standards, Vol. 67C,No. 2, pp. 161-187, 1963.

[33] Croarkin, Carroll, "Measurement Assurance Programs, Part II: Development andImplementation," NBS Special Publication 676-II, 1984.

[34] ISO, "Guide to the Expression of Uncertainty in Measurement," October 1993.

[35] Taylor, Barry N. and Chris E. Kuyatt, "Guidelines for Evaluating and Expressing theUncertainty of NIST Measurement Results," NIST Technical Note 1297, 1994 Edition,September 1994.

[36] Tan, A. and J.R. Miller, "Trend studies of standard and regular gauge block sets," Reviewof Scientific Instruments, V.62, No. 1, pp.233-237, 1991.

[37] C.F. Bruce, "The Effects of Collimation and Oblique Incidence in Length Interferometers.I," Australian Journal of Physics, Vol. 8, pp. 224-240 (1955).

[38] C.F. Bruce, "Obliquity Correction Curves for Use in Length Interferometry," J. of the OpticalSociety of America, Vol. 45, No. 12, pp. 1084-1085 (1955).

[39] B.S. Thornton, "The Effects of Collimation and Oblique Incidence in Length Interferometry,"Australian Journal of Physics, Vol. 8, pp. 241-247 (1955).

[40] L. Miller, Engineering Dimensional Metrology, Edward Arnold, Ltd., London (1962).

[41] Schweitzer, W.G., et.al., "Description, Performance and Wavelengths of Iodine StabilizedLasers," Applied Optics, Vol 12, 1973.

[42] Chartier, J.M., et. al., "Intercomparison of Northern European 127I2 - Stabilized He-Ne Lasersat λ = 633 nm," Metrologia, Vol 29, 1992.

[43] Balhorn, R., H. Kunzmann and F. Lebowsky, AFrequency Stabilization of Internal-MirrorHelium-Neon Lasers,@ Applied Optics, Vol 11/4, April 1972.

[44] Mangum, B.W. and G.T. Furukawa, "Guidelines for Realizing the International TemperatureScale of 1990 (ITS-90)," NIST Technical Note 1265, National Institute of Standards andTecnology, 1990.

[45] Edlen, B., "The Refractive Index of Air," Metrologia, Vol. 2, No. 2, 1966.

123

[46] Schellekens, P., G. WIlkening, F. Reinboth, M.J. Downs, K.P. Birch, and J. Spronck,"Measurements of the Refractive Index of Air Using Interference Refractometers,"Metrologia, Vol. 22, 1986.

[47] Birch, K.P. and Downs, M.J., "The results of a comparison between calculated and measuredvalues of the refractive index of air," J. Phys. E: Sci. Instrum., Vol. 21, pp. 694-695, 1988.

[48] Birch, K.P. and M.J. Downs, "Correction to the Updated Edlén Equation for the RefractiveIndex of Air," Metrologia, Vol. 31, 1994.

[49] C.R. Tilford, "Analytical Procedure for determining lengths from fractional fringes," AppliedOptics, Vol. 16, No. 7, pp. 1857-1860 (1977).

[50] F.H. Rolt, Gauges and Fine Measurements, Macmillan and Co., Limited, 1929.

[51] Beers, J.S. and Kang B. Lee, "Interferometric measurement of length scales at the NationalBureau of Standards," Precision Engineering, Vol. 4, No. 4, 1982.

[52] Beers, J.S., "Length Scale Measurement Procedures at the National Bureau of Standards,"NBSIR 87-3625, 1987.

[53] Beers, John S. and William B. Penzes, "NIST Length Scale Interferometer MeasurementAssurance," NISTIR 4998, 1992.

124

APPENDIX A. Drift Eliminating Designs for Non-simultaneous ComparisonCalibrations

Introduction

The sources of variation in measurements are numerous. Some of the sources are truly randomnoise, 1/f noise in electronic circuits for example. Usually the "noise" of a measurement is actuallydue to uncontrolled systematic effects such as instability of the mechanical setup or variations in theconditions or procedures of the test. Many of these variations are random in the sense that they aredescribable by a normal distribution. Like true noise in the measurement system, the effects can bereduced by making additional measurements.

Another source of serious problems, which is not random, is drift in the instrument readings. Thiseffect cannot be minimized by additional measurement because it is not generally pseudo-random,but a nearly monotonic shift in the readings. In dimensional metrology the most import cause of driftis thermal changes in the equipment during the test. In this paper we will demonstrate techniquesto address this problem of instrument drift.

A simple example of the techniques for eliminating the effects of drift by looking at two differentways of comparing 2 gauge blocks, one standard (A) and one unknown (B).

Scheme 1: A B A B

Scheme 2: A B B A

Now let us suppose we make the measurements regularly spaced in time, 1 time unit apart, and thereis an instrumental drift of ∆. The actual readings (yi) from scheme 1 are:

y1 = A (A.1a)y2 = B + ∆ (A.1b)y3 = A + 2∆ (A.1c)y4 = B + 3∆ (A.1d)

Solving for B in terms of A we get:

(A.2)

which depends on the drift rate ∆.

Now look at scheme 2. Under the identical conditions the readings are:

-Y4)-Y2-Y3+(Y121-A=B

125

y1 = A (A.3a)y2 = B + ∆ (A.3b)y3 = B + 2∆ (A.3c)y4 = A + 3∆ (A.3d)

Here we see that if we add the second and third readings and subtract the first and fourth readingswe find that the ∆ drops out:

(A.4)

Thus if the drift rate is constant - a fair approximation for most measurements if the time is properlyrestricted - the analysis both eliminates the drift and supplies a numerical approximation of the driftrate.

The calibration of a small number of "unknown" objects relative to one or two reference standardsinvolves determining differences among the group of objects. Instrumental drift, due most often totemperature effects, can bias both the values assigned to the objects and the estimate of the effectof random errors. This appendix presents schedules for sequential measurements of differences thateliminate the bias from these sources and at the same time gives estimates of the magnitude of theseextraneous components.

Previous works have [A1,A2] discussed schemes which eliminate the effects of drift forsimultaneous comparisons of objects. For these types of measurements the difference between twoobjects is determined at one instant of time. Examples of these types of measurements arecomparisons of masses with a double pan balance, comparison of standard voltage cells, andthermometers which are all placed in the same thermalizing environment. Many comparisons,especially those in dimensional metrology, cannot be done simultaneously. For example, using agauge block comparator, the standard, control (check standard) and test blocks are moved one at atime under the measurement stylus. For these comparisons each measurement is made at a differenttime. Schemes which assume simultaneous measurements will, in fact, eliminate the drift from theanalysis of the test objects but will produce a measurement variance which is drift dependent andan erroneous value for the drift, ∆.

In these calibration designs only differences between items are measured so that unless one or moreof them are standards for which values are known, one cannot assign values for the remaining"unknown" items. Algebraically, one has a system of equations that is not of full rank and needs thevalue for one item or the sum of several items as the restraint to lead to a unique solution. The leastsquares method used in solving these equations has been presented [A3] and refined [A4] in theliterature and will not be repeated in detail here. The analyses presented of particular measurementdesigns presented later in this paper conform to the method and notation presented in detail byHughes [A3].

Y3)-Y2-Y4+(Y121-A=B

126

The schemes used as examples in this paper are those currently used at NIST for gauge blockcomparisons. In our calibrations a control (check standard) is always used to generate data for ourmeasurement assurance plan [A5]. It is not necessary, however, to use a control in everymeasurement and the schemes presented can be used with any of the objects as the standard and therest as unknowns. A number of schemes of various numbers of unknowns and measurements ispresented in the appendix.

Calibration Designs

The term calibration design has been applied to experiments where only differences betweennominally equal objects or groups of objects can be measured. Perhaps the simplest such experimentconsists in measuring the differences between the two objects of the n(n-1) distinct pairings that canbe formed from n objects. If the order is unimportant, X compared to Y is the negative of Ycompared to X, there are only n(n-1)/2 distinct pairings. Of course only 1 measurement per unknownis needed to determine the unknown, but many more measurements are generally taken for statisticalreasons. Ordinarily the order in which these measurements are made is of no consequence. However, when the response of the comparator is time dependent, attention to the order is importantif one wished to minimize the effect of these changes.

When this effect can be adequately represented by a linear drift, it is possible to balance out theeffect by proper ordering of the observations. The drift can be represented by the series, . . . -3 , -2, -1 , 0, 1, 2 , 3 , . . . if there are an odd number of comparisons and by ... -5/2 , -3/2 , -1/2 , 1/2 ,3/2 , 5/2 , ... if there are an even number of comparisons.

As and example let us take n=3. If we make all possible n(n-1)=6 comparisons we get a scheme likethat below, denoting the three objects by A, B, C.

Observation Measurement

m1 A - 11/2 ∆ m2 B - 9/2 ∆

m3 C - 7/2 ∆m4 A - 5/2 ∆m5 B - 3/ 2 ∆

m6 C - 1/2 ∆ (A.5)m7 A + 1/2 ∆m8 C + 3/2 ∆m9 B + 5/2 ∆m10 A + 7/2 ∆m11 C + 9/2 ∆m12 B + 11/2 ∆

127

If we analyze these measurements by pairs, in analogy to the weighing designs of Cameron we seethat:

A B C ∆y1 = m1 - m2 = A - B - ∆ +1 - 1 -1y2 = m3 - m4 = C - A - ∆ - 1 +1 -1y3 = m5 - m6 = B - C - ∆ +1 - 1 -1 (A.6)y4 = m7 - m8 = A - C - ∆ +1 - 1 -1y5 = m9 - m10 = B - A - ∆ - 1 +1 -1y6 = m11 - m12 = C - B - ∆ - 1 +1 -1

The notation used here, the plus and minus signs, indicate the items entering into the differencemeasurement. Thus, y2 is a measurement of the difference between object C and object A.

Note the difference between the above table and that of simultaneous comparisons in reference 2 isthat the drift column is constant. It is simple to see by inspection that the drift is balanced out sinceeach object has two (+) and two (-) measurements and the drift column is constant. By extension,the effects of linear drift is eliminated in all complete block measurement schemes (those for whichall objects are measured in all possible n(n-1) combinations).

Although all schemes in which each object has equal numbers of (+) and (-) measurements is drifteliminating, there are practical criteria which must be met for the scheme to work. First, the actualdrift must be linear. For dimensional measurements the instrument drift is usually due to changesin temperature. The usefulness of drift eliminating designs depends on the stability of the thermalenvironment and the accuracy required in the calibration. In the NIST gauge block lab theenvironment is stable enough that the drift is linear at the 3 nm (0.1 µin) level over periods of 5 to10 minutes. Our comparison plans are chosen so that the measurements can be made in this period. Secondly, each measurement must be made in about the same amount of time so that themeasurements are made at fairly regular intervals. In a completely automated system this is simple,but with human operators there is a natural tendency to make measurements simpler and quicker isthe opportunity presents itself. For example, if the scheme has a single block measured two or moretimes in succession it is tempting to measure the object without removing it from the apparatus,placing it in its normal resting position, and returning it to the apparatus for the next measurement.

Finally, the measurements of each block are spread as evenly as possible across the design. Supposein the scheme above where each block is measured 4 times block, A is measured as the firstmeasurement of y1, y2, y3, and y4. There is a tendency to leave block A near the measuring pointrather than its normal resting position because it is used so often in the first part of the scheme. Thisallows block A to have a different thermal handling than the other blocks which can result in a

128

thermal drift which is not the same as the other blocks.

Restraints

Since all of the measurements made in a calibration are relative comparisons, at least one value mustbe known to solve the system of equations. In the design of the last section, for example, if one hasa single standard and two unknowns, the standard can be assigned to any one of the letters. (Thesame would be true of three standards and one unknown.) If there are two standards and oneunknown, the choice of which pair of letters to assign for the standards is important in terms ofminimizing the uncertainty in the unknown.

For full block designs (all possible comparisons are made) there is no difference which label is usedfor the standards or unknowns. For incomplete block designs the uncertainty of the results candepend on which letter the standard and unknowns are assigned. In these cases the customer blocksare assigned to minimize their variance and allow the larger variance for the measurement of theextra master (control).

This asymmetry occurs because every possible comparison between the four items has not beenmeasured. For 4 objects there are 12 possible intercomparisons. If an 8 measurement scheme isused all three unknowns cannot be compared directly to the standard the same number of times. Forexample, two unknowns can be compared directly with the standard twice, but the other unknownwill have no direct comparisons. This indirect comparison to the standard results in a slightly largervariance for the block compared indirectly. Complete block plans, which compare each block toevery other block equal number of times, have no such asymmetry, and thus remove any restrictionon the measurement position of the control.

Example: 4 block, 12 comparison, Single Restraint Design for NIST Gauge Block Calibration

The gauge block comparison scheme put into operation in 1989 consists of two standards blocks,denoted S and C, and two customer blocks to be calibrated, denoted X and Y. In order to decreasethe random error of the comparison process a new scheme was devised consisting of all 12 possiblecomparisons between the four blocks. Because of continuing problems making penetrationcorrections, the scheme was designed to use either the S or C block as the restraint and the difference(S-C) as the control parameter. The S blocks are all steel, and are used as the restraint for all steelcustomer blocks. The C blocks are chrome carbide, and are used as the restraint for chrome andtungsten carbide blocks. The difference (S-C) is independent of the choice of restraint.

We chose a complete block scheme that assures that the design is drift eliminating, and the blockscan be assigned to the letters of the design arbitrarily. We chose (S-C) as the first comparison. Sincethere are a large number of ways to arrange the 12 measurements for a complete block design, weadded two restrictions as a guide to choose a "best" design.

1. Since the blocks are measured one at a time, it was decided to avoid schemes which measured

129

the same block two or more times consecutively. In the scheme presented earlier blocks D, Aand C are all measured twice consecutively. There is a great temptation to not remove andreplace the blocks under these conditions, and the scheme assumes that each measurement ismade with the same motion and are evenly spaced in time. This repetition threatens both theseassumptions.

2. We decided that schemes in which the six measurements of each block were spread out asevenly as possible in time would be less likely to be affected by small non-linearities in the drift. For example, some schemes had one block finished its 6 measurements by the 8th comparison,leaving the final 1/3 of the comparisons with no sampling of that block.

The new scheme is as follows: S C X Y ∆

XYSXCXYSYCXSXYXCSCYXSYCS

YYYYYYYYYYY

Y

Y

12

11

10

9

8

7

6

5

4

3

2

1

111001010110110110011101010101110101011010011111001100110011

X (A.7)

When the S block is the restraint (S – L) the matrix equation to solve is:

LYX

aaXX

A'

0''

(A.8)

| a’| = [ 1 0 0 0 ] (A..9)

130

0000010120000006222002622002262102226

0''

a

aAXXCij (A.10)

00242424240200002406330240363024033602400000

2411

ijC (A.11)

Using S as the restraint, the solution to the equations is:

S = L (A.12a)

C = (1/8) (-2y1 + y2 + 2y4 + y5 - y6 - y7 + y8 - y9 - y10 + y11) + L (A.12b)

X = (1/8) (-y1 + y2 + y3 + y4 - y5 - 2y7 - y9 + y10 + 2y11 - y12) + L (A.12c)

Y = (1/8) (-y1 + 2y2 - y3 + y4 + y6 - y7 - y8 - 2y9 + y11 + y12) + L (A.12d)

∆ = (-1/12)(y1 + y2 + y3 + y4 + y5 + y6 + y7 + y8 + y9 + y10 + y11 + y12) (A.12e)

The deviations, d1, d2, ..., d12 can be determined from the equations above, or can be calculateddirectly using matrix methods. For example,

These deviations provide the information needed to obtain a value, s, which is the experiment's valuefor the short term process standard deviation, or within standard deviation σw.

131

(A.13)

The number of degrees of freedom results from taking the number of observations (n=12) less thenumber of unknowns (m=5; S, C, X, Y, ∆ ), and then adding one for the restraint. Because of thecomplete block structure (all 12 possible combinations measured) all of the standard deviations arethe same:

(A.14)

Process Control: F - Test

Continued monitoring of the measurement process is required to assure that predictions based on theaccepted values for process parameters are still valid. For gauge block calibration at NIST, theprocess is monitored for precision by comparison of the observed standard deviation, σw, to theaverage of previous values. For this purpose the value of σw is recorded for every calibration done,and is periodically analyzed to provide an updated value of the accepted process σw for each gaugeblock size.

The comparison is made using the F distribution, which governs the comparison of variances. Theratio of the variances s2 (derived from the model fit to each calibration) and σw

2 derived from thehistory is compared to the critical value F(8,∞,α), which is the α probability point of the Fdistribution for degrees of freedom 8 and ∞. For calibrations at NIST, α is chosen as 0.01 to giveF(8,∞,.01) = 2.5.

(A.15)

If this condition is violated the calibration fails, and is repeated. If the calibration fails more thanonce the test blocks are re-inspected and the instrument checked and recalibrated. All calibrations,pass or fail, are entered into the history file.

Process Control: T - Test

At NIST a control measurement is made with each calibration by using two known master blocksin each calibration. One of the master blocks is steel and the other chrome carbide. When acustomer block is steel the steel master is used as the restraint, and when a customer block is carbide,

1) + m - (n / ))) X xA( - Y(( = s 2trr

m

0 = ri

n

0 = i

2

(1/2)s=sB iiw

2.5<s=F 2t

2obs

132

the carbide master is used as the restraint. The use of a control measurement for calibrations isnecessary in order to provide assurance of the continuing accuracy of the measurements. The F-test,while providing some process control, only attempts to control the repeatability of the process, notthe accuracy. The use of a control is also the easiest method to find the long term variability of themeasurement process.

While the use of a control in each calibration is not absolutely necessary, the practice is highlyrecommended. There are systems that use intermittent tests, for example measurements of a controlset once a week. This is a good strategy for automated systems because the chance of block to blockoperator errors is small. For manual measurements the process variability, and of course theoccurrence of operator error is much higher.

The check for systematic error is given by comparison of the observed value of the differencebetween the standard and control blocks. If S is the standard it becomes the restraint, and if A isused as the control (S-A) is the control parameter for the calibration. This observed control isrecorded for every calibration, and is used to periodically used to update the accepted, or averagevalue, of the control. The process control step involves the comparison of the observed value of thecontrol to the accepted (historical) value. The comparison is made using the Student t-distribution.

The control test demands that the observed difference between the control and its accepted value beless than 3 times the accepted long term standard deviation, σt, of the calibration process. This valueof the t-distribution implies that a good calibration will not be rejected with a confidence level of99.7%.

(A.16)

The value of σt is obtained directly from the sequence of values of (S-A) arising in regularcalibrations. The recorded (S-C) values are fit to a straight line, and the square root of the varianceof the deviations from this line is used as the total standard deviation, (σt).

If both the precision (F-test) and accuracy (t-test) criteria are satisfied, the process is regarded asbeing "in control" and values for the unknown, X, and its associated uncertainty are regarded asvalid. Failure on either criterion is an "out-of-control" signal and the measurements are repeated.

The value for drift serves as an indicator of possible trouble if it changes markedly from its usualrange of values. However, because any linear drift is balanced out, a change in the value does notof itself invalidate the result.

3<)A-A(-)A-A(=T

t

acc21obs21

133

Conclusion

The choice of the order of comparisons is an important facet of calibrations, in particular if chosenproperly the comparison scheme can be made immune to linear drifts in the measurement equipment. The idea of making a measurement scheme robust is a powerful one. What is needed to implementthe idea is an understanding of the sources of variability in the measurement system. While such astudy is sometimes difficult and time consuming because of the lack of reference material aboutmany fields of metrology, the NIST experience has been that such efforts are rewarded withmeasurement procedures which, for about the same amount of effort, produce higher accuracy.

134

References

[A1] J.M. Cameron, M.C. Croarkin, and R.C. Raybold, "Designs for the Calibration of Standardsof Mass," NBS Technical Note 952, 1977.

[A2] Cameron, J.M. and G.E Hailes. "Designs for the Calibration of Small Groups of Standardsin the Presence of Drift," NBS Technical Note 844, 1974.

[A3] C.G. Hughes, III and H.A. Musk. "A Least Squares Method for Analysis of Pair ComparisonMeasurements," Metrologia, Volume 8, pp. 109-113 (1972).

[A4] C. Croarkin, "An Extended Error Model for Comparison Calibration," Metrologia, Volume26, pp. 107-113, 1989.

[A5] C. Croarkin, "Measurement Assurance Programs, Part II: Development and Implementation,"NBS Special Publication 676-II, 1984.

135

A Selection of Other Drift Eliminating Designs

The following designs can be used with or without a control block. The standard block is denotedS, and the unknown blocks A, B, C, etc. If a check standard block is used it can be assigned to anyof the unknown block positions. The name of the design is simply the number of blocks in thedesign and the total number of comparisons made.

3-6 Design 3-9 Design 4-8 Design

(One master block, (One master block, (One master block, 2 unknowns 2 unknowns, 3 unknowns, 4 measurements each) 6 measurements each) 4 measurements each)

y1 = S - A y1 = S - A y1 = S - Ay2 = B - S y2 = B - A y2 = B - Cy3 = A - B y3 = S - B y3 = C - Sy4 = A - S y4 = A - S y4 = A - By5 = B - A y5 = B - S y5 = A - Sy6 = S - B y6 = A - B y6 = C - B

y7 = A - S y7 = S - Cy8 = B - A y8 = B - Ay9 = S - B

4-12 Design 5-10 Design 6-12 Design

(One master block, (One master block, (One master block, 3 unknowns 4 unknowns, 5 unknowns, 6 measurements each) 4 measurements each) 4 measurements each)

y1 = S - A y1 = S - A y1 = S - Ay2 = C - S y2 = D - C y2 = D - Cy3 = B - C y3 = S - B y3 = E - By4 = A - S y4 = D - A y4 = E - Dy5 = A - B y5 = C - B y5 = C - Ay6 = C - A y6 = A - C y6 = B - Cy7 = S - B y7 = B - S y7 = S - Ey8 = A - C y8 = B - D y8 = A - Dy9 = S - C y9 = C - S y9 = A - By10 = B - A y10 = A - D y10 = D - Sy11 = B - S y11 = B - Ey12 = C - B y12 = C - S

136

7-14 Design 8-16 Design 9-18 Design(One master block, (One master block, (One master block, 6 unknowns 7 unknowns, 7 unknowns, 4 measurements each) 4 measurements each) 4 measurements each)

y1 = S - A y1 = S - A y1 = S - Ay2 = E - C y2 = E - G y2 = H - Fy3 = B - D y3 = F - C y3 = A - By4 = A - F y4 = D - S y4 = D - Cy5 = S - E y5 = B - E y5 = E - Gy6 = D - B y6 = G - F y6 = C - Ay7 = A - C y7 = C - B y7 = B - Fy8 = B - F y8 = E - A y8 = G - Hy9 = D - E y9 = F - D y9 = D - Sy10 = F - S y10 = C - S y10 = C - Ey11 = E - A y11 = A - G Y11= H - Sy12 = C - B y12 = D - B y12 = G - Dy13 = C - S y13 = C - S y13 = C - Sy14 = F - D y14 = G - C y14 = A - C

y15 = B - D y15 = F - Dy16 = A - F y16 = S - H

y17 = E - By18 = F - G

137

10-20 Design 11-22 Design

(One master block, (One master block, 9 unknowns 10 unknowns, 4 measurements each) 4 measurements each)

y1 = S - A y1 = S - Ay2 = F - G y2 = D - Ey3 = I - C y3 = G - Iy4 = D - E y4 = C - Hy5 = A - H y5 = A - By6 = B - C y6 = I - Jy7 = G - H y7 = H - Fy8 = I - S y8 = D - Sy9 = E - F y9 = B - Cy10 = H - I y10 = S - Ey11 = D - F y11 = A - Gy12 = A - B y12 = F - By13 = C - I y13 = E - Fy14 = H - E y14 = J - Ay15 = B - G y15 = C - Dy16 = S - D y16 = H - Jy17 = F - B y17 = F - Gy18 = C - D y18 = I - Sy19 = G - S y19 = B - Hy20 = E - A y20 = G - D

y21 = J - Cy22 = E - I

138

Appendix B: Wringing Films

In recent years it has been found possible to polish plane surfaces of hardened steelto a degree of accuracy which had previously been approached only in the finestoptical work, and to produce steel blocks in the form of end gauges which can bemade to adhere or "wring" together in combinations. Considerable interest has beenaroused by the fact that these blocks will often cling together with such tenacity thata far greater force must be employed to separate them than would be required if theadhesion were solely due to atmospheric pressure. It is proposed in this paper toexamine the various causes which produce this adhesion: firstly, showing that by farthe greater portion of the effect is due to the presence of a liquid film between thefaces of the steel; and , secondly, endeavoring to account for the force which can beresisted by such a film.

Thus began the article "The Adherence of Flat Surfaces" by H.M. Budgett in 1912 [B1], the firstscientific attack on the problem of gauge block wringing films. Unfortunately for those wishing tidysolutions, the field has not progressed much since 1912. The work since then has, of course, addedmuch to our qualitative understanding of various phenomena associated with wringing, but there isstill no clear quantitative or predictive model of wringing film thickness or its stability in time. Inthis appendix we will only describe some properties of wringing films, and make recommendationsabout strategies to minimize problems due to film variations.

Physics of Wringing Films

What causes wrung gauge blocks to stick together? The earliest conjectures were that sliding blockstogether squeezed the air out, creating a vacuum. This view was shown to be wrong as early as 1912by Budgett [B1] but still manages to creep into even modern textbooks [B2]. It is probable thatwringing is due to a number of forces, the relative strengths of which depend on the exact nature ofthe block surface and the liquid adhering to the surface. The known facts about wringing aresummarized below.

1. The force of adhesion between blocks can be up to 300 N (75 lb). The force of theatmosphere, 101 KPa (14 psi), is much weaker than an average wring, and studies have shownthat there is no significant vacuum between the blocks.

2. There is some metal-metal contact between the blocks, although too small for a significantmetallic bond to form. Wrung gauge blocks show an electrical resistance of about 0.003Ω [B3]that corresponds to an area of contact of 10-5 cm2.

3. The average wringing film thickness depends on the fluid and surface finishes, as well as theamount of time blocks are left wrung together. Generally the thickness is about 10 nm (0.4 µin),but some wrings will be over 25 nm (1 µin) and some less than 0. (Yes, less than zero.)[B3,B4.B5,B6]

139

4. The fluid between blocks seems to provide much of the cohesive force. No matter how ablock is cleaned, there will be some small amount of adsorbed water vapor. The normalwringing procedure, of course, adds minute amounts of grease which allows a more consistentwringing force. The force exerted by the fluid is of two types. Fluid, trapped in the very smallspace between blocks, has internal bonds that resist being pulled apart. The fluid also has asurface tension that tends to pull blocks together. Both of these forces are large enough toprovide the observed adhesion of gauge blocks.

5. The thickness of the wringing film is not stable, but evolves over time. First changes are dueto thermal relaxation, since some heat is transferred from the technician's hands during wringing. Later, after the blocks have come to thermal equilibrium, the wring will still change slowly. Over a period of days a wring can grow, shrink or even complicated combinations of growth andshrinkage [B5,B6].

6. As a new block is wrung repeatedly the film thickness tends to shrink. This is due tomechanical wear of the high points of the gauge block surface [B5,B6].

7. As blocks become worn and scratched the wringing process becomes more erratic, until theydo not wring well. At this point the blocks should be retired.

There may never be a definitive physical description for gauge block wringing. Besides the papersmentioned above, which span 60 years, there was a large project at the National Bureau of Standardsduring the 1960's. This program studied wringing films by a number of means, includingellipsometry [B8]. The results were very much in line with the 7 points given above, i.e., on apractical level we can describe the length properties of wringing films but lack a deeperunderstanding of the physics involved in the process.

Fortunately, standards writers have understood this problem and included the length of one wringingfilm in the defined block length. This relieves us of determining the film thickness separately sinceit is automatically included whenever the block is measured interferometrically. There is someuncertainty left for blocks that are measured by mechanical comparison, since the length of themaster block wringing film is assumed to be the same as the unknown block. This uncertainty isprobably less than 5 nm ( .2 µin) for blocks in good condition.

REFERENCES

[B1] "H.M. Budgett, "The Adherence of Flat Surfaces," Proceedings of the Royal Society, Vol.86A, pp. 25-36 (1912).

[B2] D.M. Anthony, Engineering Metrology, Peragamon Press, New York, N.Y. (1986).

[B3] C.F. Bruce and B.S. Thornton, "Adhesion and Contact Error in Length Metrology," Journal

140

of Applied Physics, Vol. 17, No.8 pp. 853-858 (1956).

[B4] C.G. Peters and H.S. Boyd, "Interference methods for standardizing and testing precisiongauge blocks," Scientific Papers of the Bureau of Standards, Vol. 17, p.691 (1922).

[B5] F.H. Rolt and H. Barrell, "Contact of Flat Surfaces," Proc. of the Royal Society (London),Vol. 106A, pp. 401-425 (1927).

[B6] G.J. Siddall and P.C. Willey, "Flat-surface wringing and contact error variability," J. ofPhysics D: Applied Physics, Vol. 3, pp. 8-28 (1970).

[B7] J.M. Fath, "Determination of the Total Phase, Basic Plus Surface Finish Effect, for GaugeBlocks," National Bureau of Standards Report 9819 (1967).

141

Appendix C. Phase Shifts in Gauge block Interferometry

When light reflects from a surface there is a shift in phase; the reflected light appears to havereflected from a plane below the mechanical surface, as shown in figure C. Many years ago, whensurface finishes were rougher, the main component of phase differences between metal surfaces wasdue to surface finish. Current manufacturing practice has reduced this component to a small part ofthe phase shift. There are still significant phase shift differences between blocks of differentmanufacturers and different materials.

Nonconducting materials, such as quartz and glass have phase shifts of 180 º. Electrical conductorswill have phase shifts less than 180º by 10º to 30º. The theoretical foundation for this shift is givenin Ditchburn [C1]. From an understanding of electromagnetic behavior of electrons near the metalsurface the phase shift could be calculated. There have been such calculations using the Drudemodel for electrons. Unfortunately this classical model of electrons in metals does not describeelectron behavior at very high frequencies, and is only useful for calculations of phase shifts forwavelengths in the infrared and beyond. There have been no successful calculations for phase shiftsat higher (visible light) frequencies, and there is no currently available theory to predict phase shiftsfrom any other measurable attributes of a metal.

Given this fact, phase shifts must be measured indirectly. There is one traditional method to measurephase shift, the "slave block" method [C2]. In this method an auxiliary block, called the slave block,is used to help find the phase shift difference between a block and a platen. The method consists oftwo steps, shown schematically in figure C2 and C3.

Figure C. The interferometric length, Ltest, includes the mechanical length, the wringing filmthickness and the phase change at each surface.

test

platen

block

mechL

L

L

wring

142

Step 1. The test and slave blocks are wrung down to the same platen and measured independently. The two lengths measured consist of the mechanical length of the block, the wringing film and thephase changes at the top of the block and platen, as in C.

Figure C2. Measurements for determining the phase shift difference between a blockand platen by the slave block method.

The general formula for the measured length of a wrung block is:

Ltest = Lmechanical + Lwring + Lplaten phase - Lblock phase (C.1)

For the test and slave blocks the formulas are:

Ltest = Lt + Lt,w + (φplaten-φtest) (C.2)

Lslave = Ls + Ls,w + (φplaten-φslave) (C.3)

Step 2. Either the slave block or both blocks are taken off the platen, cleaned, and rewrung as a stackon the platen. The new length measured is:

Ltest + slave = Lt + Ls + Lt,w + Ls,w + (φplaten-φslave) (C.4)

If this result is subtracted from the sum of the two previous measurements we find that:

Ltest + slave - Ltest - Lslave = (φtest-φplaten) (C.5)

Test SlaveSlave

Test

143

The weakness of this method is the uncertainty of the measurements. The uncertainty of onemeasurement of a wrung gauge block is about 30 nm (1.2 µin). Since the phase measurementdepends on the difference between 3 measurements, the phase measurement is uncertain to about 3x uncertainty of one measurement, or about 50 nm. Since the phase difference between block andplaten is generally about 20 nm, the uncertainty is larger than the effect. To reduce the uncertaintya large number of measurements must be made, generally between 50 and 100. This is, of course,very time consuming.

Table C shows a typical result of the slave block method for blocks wrung to quartz. The uncertaintyis, unfortunately large; we estimate the standard uncertainty (95% confidence level) to be 8 nm.

Table C

Corrections to Nominal Lengths and Phase in Nanometers

Manufacturer Material Measured Phase(nanometers)

DoAll steel 25C.E. Johansson steel 50Matrix steel 50Pratt & Whitney steel 50Webber steel 55Webber chrome carbide 30Mitutoyo steel 40

We have made attempts to measure the phase for the ceramic (zirconia) gauge blocks but have hadproblems with residual stresses in the block stack. We have measured the phase to be 35 nm, butthe geometry leads us to suspect that the measured phase might be a distortion of the block ratherthan the interferometric phase.

There have been a number of attempts to measure phase by other means [C3], but none have beenparticularly successful. One method, using Newton's rings formed by a spherical quartz surface incontact with a metal surface is under investigation [C4]. With computer vision systems to analyzefringe patterns, this method may become useful in the future.

As a rule, blocks from a single manufacturer have the same phase since the material and lappingmethod (and thus surface finishes) are the same for all blocks. Blocks from the same manufacturerbut with different materials, or the same material from different manufacturers, generally havedifferent phases. One way to reduce phase problems is to restrict your master blocks to onemanufacturer and one material.

Another way to reduce effects of phase change is to use blocks and platens of the same material andmanufacturer. While this seems simple in principle, it is not as easy in practice. Most manufacturers

144

of gauge blocks do not make gauge block platens, and even when they do, lapping procedures forlarge platens are not always the same as procedures for blocks. At NIST we currently do not haveany block/platen combinations which have matched phase.

REFERENCES

[C1] R.W. Ditchburn, Light, Interscience Publications, 2nd. edition, 1963.

[C2] Miller, L., Engineering Dimensional Metrology, Edward Arnold, Ltd., London, 1962.

[C3] J.M. Fath, "Determination of the Total Phase, Basic Plus Surface Finish Effect, for GaugeBlocks," National Bureau of Standards Report 9819 (1967).

[C4] K.G. Birch and R. Li, "A Practical Determination of the Phase Change at Reflection,"IMEKO Symposium on Laser Applications in Precision Measurement, pp. 5-25, Budapest,1986.

145

Appendix D. Deformation Corrections

Whenever two materials are forced into contact there is an elastic deformation. For gauge blockcomparators the contact is between a spherical probe and the flat gauge block surface. The followingformula, from a CSIRO Technical Note [D1] has been found to be identical to the earlier NBSnomographs [D2]. The nomographs were developed to avoid the cumbersome calculations neededfor deformation corrections. However, microcomputers have made the formula easier to use thannomographs, and we therefore give only the deformation formula.

(D.1)

α is the deformation in millimeters

P is the force in NEWTONS,

D is the diameter of the probe in MILLIMETERS,

Vi is a number characteristic of the probe or block material from the table below:

Material V (10-8)

Steel 139 Chrome Carbide 86Tungsten Carbide 40Ceramic 139Diamond 43

Example 1: A steel gauge block is measured with a two point probe system each with a sphericaldiamond contact of 6 mm diameter. The top force is 1 N and the bottom force is 1/3 N.

Top Contact: D = 6 mmF = 1 NV1= 139 x 10-8 steel blockV2= 43 x 10-8 diamond sphere

D )V + V( P 2.231 31-3

2

2132

146

α = 2.231 * (1)2/3 * (139 x 10-8 + 43 x 10-8)2/3 * (6)-1/3 (D.2)

= 0.000181 mm = 0.181 µm

Bottom Contact: D = 6 mmF = 1/3 NV1= 139 x 10-8 steel blockV2= 43 x 10-8 diamond sphere

α = 2.231 * (1/3)2/3 * (139 x 10-8 + 43 x 10-8)2/3 * (6)-1/3 (D.3)

= 0.000087 mm = 0.087 µm

Thus the total deformation is 0.00027 mm or 0.27 µm.

Example 2: Same apparatus but measuring a chrome carbide gauge block:

Top Contact: D = 6 mmF = 1 NV1= 86 x 10-8 chrome carbide blockV2= 43 x 10-8 diamond sphere

α = 2.231 * (1)2/3 * (86 x 10-8 + 43 x 10-8)2/3 * (6)-1/3 (D.4)

= 0.000145 mm = 0.145 µm

Bottom Contact: D = 6 mmF = 1/3 NV1= 86 x 10-8 chrome carbide blockV2= 43 x 10-8 diamond sphere

α = 2.231 * (1/3)2/3 * (86 x 10-8 + 43 x 10-8)2/3 * (6)-1/3 (D.5)

= 0.000070 mm = 0.070 µm

147

Thus the total deformation is 0.00022 mm or 0.22 µm.

If we were to use a steel master block to calibrate a chrome carbide block under the above conditionswe see that the penetration correction for steel and chrome carbide blocks differ substantially. Thetotal length correction of the calibrated block is given by:

Lu = Mu - Ms + Ls + (αs - αu) (D.6)

where Lu is the calibrated length of the unknown block, Mu is the measured value of the unknownblock, Ms is the measured value of the reference standard, and αs and αu are penetration correctionsof the standard and unknown respectively, and αs-αu=0.05 µm (2.0 µin).

REFERENCES

[D1] Puttock, M.J. and E.G. Thwaite, "Elastic Compression of Spheres and Cylinders at Point andLine Contact," Natoinal Standards Laboratory Technical Paper No. 25, CSIRO, 1969.

[D2] Beers, J.S. and J.E. Taylor. "Contact deformation in Gauge Block Comparisons," NBSTechnical Note 962, 1978.